Noise background, apparatus for processing noise background, method of providing noise background parameters, method of providing spectral representation of audio signal noise background, computer program and encoded audio signal

FIELD: physics, acoustics.

SUBSTANCE: noise filler for creating a noise-filled spectral representation of an audio signal based on an input spectral representation of the audio signal consists of a spectral region identifier configured to identify spectral regions of the input spectral representation spaced from non-zero spectral regions of the input spectral representation by at least one intermediate spectral region, to obtain identified spectral regions, and a noise inserter configured to selectively introduce noise into the identified spectral regions to obtain the noise-filled spectral representation of the audio signal. A noise filling parameter calculator for calculating a noise filling parameter based on a quantised spectral representation of an audio signal comprises a spectral region identifier, as mentioned above, and a noise value calculator configured to selectively consider quantisation errors in the identified spectral regions for calculation of the noise filling parameter. Accordingly, an encoded audio signal representation representing the audio signal can be obtained.

EFFECT: improving noise-filling of an encoded audio signal while limiting undesirable distortions.

15 cl, 9 dwg

Background of invention

Implementation of the invention allows the use of homosapien with the purpose of creation is filled with noise spectral representation of the audio signal based on the input spectral representation of the audio signal, to calculate a parameter sumusuporta on the basis of the quantized spectral representation of the audio signal to create an encoded audio signal, use a new method of creation is filled with noise spectral audio signal, to use the new calculation method parameter homoserine based on the quantized spectral audio signal, use a computer program for implementing the above methods.

Hereinafter will be described some scenarios in which the invention may be implemented. Many encoders audio signals in the frequency domain based on the idea that certain areas of the frequency or range (for example, the frequency of the line or spectral lines obtained by converting the time domain into the frequency domain) is more important than the other spectral region. In connection with this spectral region with high psychoacoustic relevance are usually encoded with greater precision than spectral region with less psychoacoustic relevance. Psychoacoustic relevance R is EIT spectral regions can be, for example, calculated using a psychoacoustic model, which takes into account the masking weaker spectral regions adjacent stronger spectral peaks.

If you need to lower the bitrate of the encoded audio signal to a lower level, some spectral region quanthouse with very low precision (e.g., with an accuracy of one bit or two bits). Accordingly, many spectral region, quantized with low precision, quanthouse to zero. Therefore, at low bit rate audio coders with conversion inevitably operate with distortion, especially with the distortion caused by the frequency lines quantized to zero. Indeed, the coarse quantization of the spectral values at low bitrate audio coding can lead to very razrezhennogo spectrum after inverse quantization, as many spectral lines can be quantized to zero. These frequency gaps in the recovered signal to produce undesirable audible distortion. This can make the reproduced sound is too sharp or unstable (whistling high tone), when the frequency gaps in the spectrum moves from frame to frame.

Sumusuporta is a means to disguise these distortions by filling on the side of the decoder, the quantized to n is La ratios or frequency bands of random noise. The energy of the inserted noise is determined by the parameter, calculated and transmitted by the encoder.

There are various concepts homoserine. For example, the so-called AMR-WR+ combines sumusuporta and discrete Fourier transform (DFT), as described, for example, in reference [1]. In addition, the international standard ITU-T G.729.1 defines the concept, which combines sumusuporta and change the discrete cosine transform (MDCT). Details are provided in [2].

Further aspects of homoserine described in the international patent application PCT/IB2002/001388 from Koninklijke Philips Electronics NV (see reference [3]).

However, traditional approaches to samozabvennuyu lead to significant sound distortion.

This raises the need to establish the concept of homoserine, which will improve your audio experience.

Brief description of the invention

The form of implementation of the present invention is homosapien to get filled with noise spectral audio signal based on the input spectral audio signal. Homosapien includes the ID of the spectral region to be used for identifying spectral regions (e.g., spectral lines or spectral bins) input spectral signal, the Department is tion from non-zero spectral regions (e.g., spectral lines or spectral bins) input spectral signal, at least one intermediate spectral region to obtain the identified spectral regions. Homosapien also includes a device for inserting noise, designed to selectively insert noise in the identified spectral region (for example, a spectral line or spectral bins) to get filled with noise spectral representations of the audio signal.

The application of this invention is based on the fact that tonal spectral components of the audio signal, usually worse from the point of view of auditory impressions, if sumusuporta used in close proximity to these tonal components. In this regard, it is established that improve the listening experience samozabvennoi audio signal can be achieved if you apply sumusuporta only in spectral regions that are distant from such tonal, non-zero spectral regions. Accordingly, tonal spectral components of the audio signal (which are not quantized to zero quantized spectral representation of the input in homosapien) remain audible (i.e. not washed away in close proximity to noise), thus it is possible to effectively avoid the greater the x, the spectral gaps.

In a preferred embodiment of the present invention, the ID of the spectral range is designed to determine the spectral lines of the input spectral representation, which are quantized to zero and which include at least a first specified number of low-frequency neighboring spectral lines quantized to zero, and at least a second specified number of high-frequency neighboring spectral lines quantized to zero, as a certain spectral region, where the first predefined number is greater than or equal to one and where the second predefined number is greater than or equal to one. In this way the implementation of the invention, the device for inserting noise is designed to selectively introduce noise in certain spectral lines, leaving spectral lines quantized to non-zero values, and the spectral lines quantized to zero, but not having the first specified number of low-frequency neighboring spectral lines quantized to zero, or the second specified number of high-frequency neighboring spectral lines quantized to zero, not subjected to samozabvennuyu. Thus, sumusuporta selective in the sense that the noise is introduced only in the spectral lines quantized to zero and which are at a distance from the lines quantized to non-zero values, as in Rheem, and at the bottom of the spectral direction, for example, the first predetermined number of low-frequency neighboring spectral lines quantized to zero, and the second predetermined number of high-frequency neighboring spectral lines quantized to zero.

In a preferred embodiment of the invention, the first predefined number is equal to the second specified number, such that the minimum distance in the direction of the frequency up lines quantized to non-zero values, the minimum distance in the direction of the frequency down from lines quantized to non-zero values.

In a preferred implementation of the invention homosapien is designed to make noise only in the spectral region in the upper part of the spectral representation of the audio signal, leaving the lower part of the spectral representation of the audio signal is not affected by sumusuporta. This concept is useful because, as a rule, the higher frequencies are less important for auditory perception than low frequencies. Values quantized to zero, mainly occur in the second half of the spectrum (i.e. high frequencies). Adding noise at high frequencies are less likely to lead to the ultimate restoration of the sound.

In a preferred embodiment of the invention, the identifier of the spectral region summeru the t quantized intensity values (for example, the values of the energy or amplitude) in the spectral regions in a given bilateral spectral surrounded by this spectral region (i.e. spectral environment, covering low and higher frequencies) for General values and estimate the total value in order to decide whether a given spectral region identified spectral region or not. It was found that the sum of the values of the quantized energy spectrum in bilateral spectral surrounded by this spectral region is important for the decision whether to apply sumusuporta in this spectral region.

In another preferred embodiment of the invention the identifier spectral range is designed to scan a range of spectral regions of the input spectral representation for the detection of related sequences spectral regions, quantized to zero, and determining one or more Central spectral regions (i.e., not lying on the boundary spectral region) above the adjacent sequences as identified spectral regions.

It was found that the identification of a specific "Run - length " in the spectral regions, quantized to zero, is a task that is characterized by a particularly low computational SL is a possibility. To identify such related sequences spectral regions can be determined whether all spectral regions within the sequence of spectral bands is quantized to zero, which can be calculated using a relatively simple algorithm or circuit. If it is determined that such contiguous sequence of spectral regions quantized into zero, one or more of the internal spectral sequence (which are relatively far from the spectral regions outside the sequence of spectral regions) are identified spectral region. Thus, using a range scan spectral regions (for example, choosing shifted sequence of spectral bands), it is possible to carry out an effective analysis of the spectral representation to determine spectral region, quantized to zero and distant from spectral regions, quantized to non-zero value, the specified minimum distance.

Another use of the invention is the use of computer parameter homoserine to calculate the parameter sumusuporta on the basis of the quantized spectral representation of the audio signal. The calculation of the parameter homoserine including the AET in itself identifier spectral region, used to identify spectral regions of the quantized spectral representation, separated from non-zero spectral regions of the quantized spectral representation, at least one intermediate spectral region to identify the identified spectral regions. The calculation of the parameter homoserine includes the transmitter jitter values, which is intended for selective consideration of quantization errors in the identified spectral regions to calculate the parameter homoserine. The evaluator parameter homoserine based on the idea that it is desirable to limit samozaparcia from the decoder spectral regions, which are separated from the tonal spectral regions (quantized to non-zero value), and that, taking this concept into consideration, therefore, the parameter of the noise must be calculated on the side of the encoder. Accordingly, it is possible to get a parameter sumusuporta that is particularly well suited for the above-described concept of the decoder. In addition, it was found that spectral region, which kwantowani to zero, but are very close to the spectral regions of the quantized to a non-zero value, often do not reflect a truly noise-like audio content, but rather, very closely associated with the neighboring tone is Linyi (quantized to non-zero value) spectral regions. Accordingly, it was found that, as a rule, it is undesirable to consider the quantization errors of the spectral regions that are close to the spectral regions, quantized to non-zero value, to calculate a parameter homoserine, because it usually leads to a strong overestimation of the noise, resulting in a too noisy reconstructed spectral representation.

Thus, the concept of calculating a parameter homoserine described here can be used in combination with the above described concept of homoserine and even in combination with the usual concepts of homoserine.

In preferred embodiments, use of the invention the concept of the identification of spectral regions, which was discussed in connection with sumsemanntill, can also be used in combination with a calculation parameter homoserine.

In a preferred embodiment, use of the invention, the transmitter noise values is to consider the actual energy quantization errors are identified spectral region for calculation of the parameter homoserine. It was found that consideration of the actual quantization errors (not approximate the quantization error or the average quantization errors), as a rule, leads to higher results, so it is to the actual quantization error is usually different from the average quantization errors.

In the preferred use case, the transmitter noise values are intended to determine the energy of the error naturalnego quantization, distributed on many of the identified spectral regions in relation to energy errors of pitch quantization, concentrated in a single spectral region. This concept is based on the discovery of the fact that nationally broadband noise, the average energy is below the quantization threshold and which as a consequence of the quantized to zero, has a much greater value for sumasampalataya than one tone sound component, the intensity of which is below the quantization threshold, even if nationally broadband noise and tonal component were quantized to zero. The reason is that homosapien by generating a random noise at the decoder can simulate the missing nationally broadband noise in the quantized spectral representation, but not missing the tonal component. Thus, the preference naturalnych components of noise, quantized to zero, before the tonal components quantized to zero, entails a more realistic sound reconstruction. This is also due to the fact that the auditory perception of a person deteriorates due to the presence of the spectral gap is much larger (e.g. the measures in the absence of broadband noise, quantized to zero)than due to the lack of small spectral peak, quantized to zero. Tonal components can be concentrated in a single spectral line or can be distributed over several adjacent spectral lines (e.g., i-1, i, i+1). The spectral region may, for example, consist of one or more spectral lines.

In the preferred use case, the transmitter noise values is to calculate the total energy value of the error of the logarithmic quantization in the identified spectral regions to obtain parameter homoserine. By calculating the total energy values of the errors of the logarithmic quantization in the identified spectral regions described above relative superiority naturalnych spectral regions, quantized to zero, over tonal areas, quantized to zero, you can get the most effective manner.

Another possible application of the invention involves creating a representation of the encoded audio signal to represent the audio signal. The representation of the encoded audio signal includes a representation of the encoded quantized spectral range of the audio signal and the encoded parameter samozamenyaemyh homoserine represents an error of quantization of spectral bands spectral representation of the domain quantized to zero, and distant from the spectral representation of the spectral domain, quantized to non-zero value, the specified number of intermediate spectral regions. The above representation of the encoded audio signal is used to sumasampalataya described above, and can be obtained using the solver option homoserine mentioned above. The representation of the encoded audio signal allows the reconstruction of the audio signal with a particularly good sound quality, because the parameter homoserine selectively reflects the quantization error of the quantized representation of the spectral domain for these spectral regions where there is significant noise information, which should be selectively considered for homoserine from the decoder.

Another use of the invention makes it possible to create a method of the view is filled by the noise of the audio signal.

Another option is the use of the invention makes it possible to create a method of calculating the parameter sumusuporta on the basis of the quantized spectral representation of the audio signal.

Another option is the use of the invention makes possible the creation of a computer program for implementing the above methods.

A brief description of the drawings.

Applications of the invention will be further described with reference to the attached drawings:

Figure 1 shows a block diagram of sumasampalataya depending on the different applications of the invention;

Figure 2 shows the block diagram of the decoder of the audio signal containing homosapien in accordance with this invention;

Figure 3 shows the program pseudo-code to implement the functionality of sumasampalataya figure 1;

Figure 4 shows a graphical representation of the identification of spectral regions that can be performed in samozaodrasle figure 1;

Figure 5 shows the block diagram of the calculation parameter homoserine, in accordance with a variant of the invention;

Figure 6 shows the program code to implement the functionality of the calculation parameter homoserine in accordance with Figure 5;

7 shows a block diagram of a method of obtaining a spectral representation with sumusuporta for the audio signal based on the input spectral representation of the audio signal;

On Fig shows the block diagram of the method of calculation of the parameter sumusuporta on the basis of the quantized spectral representation of the audio signal;

Figure 9 shows a graphical representation of the representation of the audio signal in accordance with a variant applied the I invention.

Homosapien figure 1-4

Figure 1 shows a block diagram of sumasampalataya 100 in accordance with a variant application of the invention. Homosapien 100 is designed to receive the input spectral representation 110 of the audio signal, for example, in the form of decoded spectral coefficients (which can be, for example, quantized or inversely quantized). Homosapien 100 is also designed to create a filled spectral representation 112 of the audio signal based on the input spectral representation 110.

Homosapien 100 contains the ID of the spectral region 120, which is designed to identify spectral regions of the input spectral representation 110, separated from non-zero spectral regions of the input spectral representation 110, at least one intermediate spectral region for information 122 indicating the identified spectral region. Homosapien 100 also includes an input device noise 130, which is designed to selectively introduce noise in the identified spectral region (described information 122) for receiving samozabvennoi spectral representation 112 of the audio signal.

As for functionality sumasampalataya 100, then, in General, we can say that Somoza Omnitel 100 selectively fills noise spectral region (for example, spectral line or spectral bin of the input spectral representation 110, for example, by replacing the spectral values of the spectral lines quantized to zero, the spectral values describing the noise. Thus, the spectral holes or spectral gaps in the input spectral representation 110 can be filled, they can, for example, be caused by coarse quantization of the input spectral representation 110. However, homosapien 100 does not make the noise all the spectral lines quantized to zero (i.e. spectral lines, the spectral values are quantized to zero). On the contrary, homosapien 100 only makes the noise in such spectral lines quantized to zero, which are at a sufficient distance from any spectral lines quantized to non-zero value. Thus, sumusuporta does not completely fill the spectral gaps or spectral gaps, and stores the spectral distance of at least one spectral region (or, at least, the distance to any other specified number of spectral bands) between the spectral lines, which made noise and spectral lines quantized to non-zero value. Thus, the spectral distance between the filling noise introduced into the spectral representation, and SpectraLine lines, quantized to a non-zero value is stored in such a way that significant for psychoacoustics spectral lines (which are not quantized to zero in the input spectral representation of the audio signal) can be visible (due to the spectral distance in a specified number of one or more spectral regions) from noise fill, made in the range sumsemanntill. Accordingly, the most significant psycho-acoustic audio content (represented spectral lines with non-zero values in the input spectral representation 110) can be clearly distinguished because of the large spectral gaps can be avoided. This is because sumusuporta selectively disconnects near the spectral lines of the input spectral representation of the quantized to a non-zero value, while sumusuporta occurs in the Central regions of spectral space or spectral gaps.

Next will be described the conditions of use for sumasampalataya 100 with reference to Figure 2. Figure 2 shows the block diagram of the decoder of the audio signal 200 in accordance with a variant of use of the invention. Decoder audio signal 200 includes as a key component of homosapien 100. Decoder audio signal 200 also includes a decoder spectral coefficients is and 210, which is designed to obtain perceptions of the encoded sound signal 212 and generate decoded optional inversely quantized representation 214 of the spectral coefficients of the encoded sound signal. The decoder spectral coefficient 210 may include, for example, the entropy decoder (for example, the arithmetic decoder or the decoder length of the series) and, if necessary, an inverse quantizer for outputting the decoded representation 214 of spectral coefficients (for example, in the form of an inversely quantized coefficients) of the representation of the encoded audio signal 212. Homosapien 100 is designed to receive the decoded representation 214 of spectral coefficients (which are not necessarily Vice versa) as the input spectral representation 110 of the audio signal.

Decoder audio signal 200 also includes a selector noise factor 220, which is designed to extract the noise factor 222 of the representation of the encoded audio signal 212 and transmit the extracted noise factor 222 to sumasampalataya 100. Decoder audio signal 200 also includes a spectrum shaper 230, which is designed to obtain a reconstructed representation of the spectrum 232 from sumasampalataya 100. The reconstructed representation of the spectrum 232 may be, for example, is Shu who zapolneniya spectral representation 112, received from sumasampalataya.

The spectrum shaper 230, which can be regarded as optional, is designed to obtain information about the spectrum 234 based on a reconstructed representation of the spectrum 232. Decoder audio signal 200 further comprises a Converter spectral domain into the time domain 240, which receives the representation of the spectrum 234, obtained from spectrum shaper 230, or in the absence of a spectrum shaper 230 reconstructed representation of the spectrum 232, and based on this you can get an idea of the sound signal in the time domain 242. Converter spectral domain into the time domain 240 may be, for example, is designed to perform the inverse discrete cosine transform (IMDCT).

In a preferred embodiment, sumusuporta on the side of the decoder includes the following steps (or should be in the following stages).

1. Decoding a minimum level of noise.

2. Decoding the quantized values of the frequency lines.

3. The determination of the spectral regions in the selected portion of the spectrum, where the length of a series of zeros is higher than the minimum length of the series.

4. The use of randomly generated character to the decoded noise level for each of the lines in certain areas.

The minimum noise level decoderules is as follows:

nf_decoded=0.0625 * (8-index).

The detected spectral region, for example, are chosen in the same way as is done on the side of the encoder (as described below).

Gaussian noise without storing data in MDCT field is generated by a spectrum with the same amplitude for all lines, but with random characters. Thus, for each of the lines within the selected areas, the decoder generates a random sign (-1 or +1) and applies it to the decoded own noise. However, it can also be applied to other methods of inserting noise.

Further, should a more detailed description with reference to figures 1, 2, 3 and 4, where Figure 3 shows the program pseudocode for algorithm homoserine from the decoder that may be implemented by sumsemanntill 100, and where Figure 4 shows a graphical representation of homoserine.

Let's start with the fact that the decoding self-noise can be performed by extracting the noise factor 220, which receives, for example, the estimated value of noise (also briefly referred to as "the index") and issues on the basis of the decoded value of the noise factor 222 (also referred to as nf_decoded"). Estimated value of the noise may be, for example, is encoded using three or four bits, and it may be, for example, an integer in the range from 0 to 7, or an integer in the range from 0 to 15.

Quantized mn the treatment frequency lines (also called "spectral line or spectral bins") can be obtained from the decoder spectral coefficient 210. Accordingly, the resulting quantized (or, if necessary, back the quantized values of the spectral lines (also called "spectral coefficients), which are referred to as "quantized (x(i))". Here i denotes the index of the frequency values of the spectral lines.

Subsequently, the spectral range detected by sumsemanntill 100 in the selected part of the spectrum (for example, in the upper part of the spectrum starting at the specified index of the frequency of spectral lines (i), where the length of the series of zeros (i.e. the values of the quantized spectral lines quantized to zero) is higher than the minimum length of the series. The detection spectral region is of the first block part 310 of the algorithm 300 figure 3. As can be seen from the first part/the first node 310 from the algorithm 300, the set R of detected regions is initialized/is in original condition/is set to the initial state/calibrated as the empty set at the beginning of the algorithm (R={};).

In the case of the algorithm in figure 3, the minimum length of a series is set at a fixed value of 8, but, of course, any other value may be selected.

Subsequently, therefore, thus, for many of the considered spectral lines (indicated by the current variable index "lines") determines whether each of the considered spectral lines is Vostochnyy environment of spectral lines, quantized to zero (and is itself considered spectral line quantized to zero). For example, all the lines in the second half of the spectra can be considered sequentially, with the line under consideration at the moment, is appointed by the frequency index "index-line". For the considered line, indicated by an index line, calculates the sum of the quantized spectral coefficients quantized (x(i))" in the environment starting from the index of the spectral frequency line index line (MinimalRunLength)/2" to index the spectral frequency line index line+MinimalRunLength)/2". If it is determined that the sum of the values of the spectral lines in the environment under consideration at the moment of the spectral line (with an index of frequency of the spectral line index line) is equal to zero, then under consideration at the moment of the spectral line (or, more precisely, the index of the frequency of the spectral line index line) is added to the set R of detected regions (or detected spectral lines). Therefore, if the index of the frequency of the spectral line is added to the set R, this means that the spectral line index lines between the index lines - MinimalRunLength)/2 and index lines + MinimalRunLength)/2" all contain the values of the spectral lines quantized to zero.

Accordingly, in the first part 310 of the program pseudo-code 310 is obtained sets the R indexes of the frequencies of spectral lines "index line", which contains those (and only those) of the spectral line under consideration of the spectral parts that are "sufficiently" far away (i.e. at least MinimalRunLength/2 lines) from any spectral lines quantized to non-zero value.

The discovery of this region is represented by Figure 4, which shows a graphical representation 400 of the spectrum. The abscissa 410 shows the frequency of spectral lines using the index of the frequency of spectral lines "index line". The ordinate 412 shows the intensity (i.e., amplitude and energy) spectral lines. As shown in figure 4, the part of the spectrum, illustrates a graphical representation 400, consists of four spectral lines 420, 420b, 420s and 420d, quantized to non-zero value. In addition, between the spectral lines 420s and 420d there are 11 spectral lines 422a-422k, quantized to zero. In addition, it is assumed that the spectral line is considered to be sufficiently distant from the spectral lines quantized to a non-zero value only if there are at least four spectral lines quantized to zero, between the observed spectral line and any other spectral line, quantized to non-zero value (and, of course, if the considered spectral line itself is quantized to zero). However, when considering the spectral line a, discovered that spectral the Naya line a adjoins the spectral line s, which is not quantized to zero, thus, the frequency index of the spectral line a is not part of a set R, calculated by the first part 310 of the algorithm 300. In addition, it is found that the spectral line 422b, s, 422d, and far enough away from the spectral lines quantized to non-zero value, so that the indexes of the frequencies of spectral lines from 422b to 422d also may not be part of the set R. In contrast, it should be noted that the spectral line e separated far enough from the spectral lines quantized to non-zero value, because the spectral line a is the line center (or, in General, the Central line), in the sequence of 9 adjacent spectral lines quantized to zero. Thus, the frequency index of the spectral line e will be part of the set R, calculated in the first part 310 of the algorithm 300. The same is true for the spectral lines 422f and 422g, since the indexes of the frequencies of spectral lines 422f and 422g will be part of the set R, which is defined in the first part 310 of the algorithm 300, because of the spectral line 422f, 422g are relatively far from the lower-frequency spectral lines 420, 420b and 420s, quantized to non-zero value, and any higher frequency spectral lines quantized to non-zero value. On the other hand, the spectral line 422h, 422i,422j and C will not be part of the set R, because these spectral lines are too close from the point of view of frequency, in addition, the spectral line 420d quantized to non-zero value.

Thus, R does not contain the indexes of the frequencies of spectral lines 420, 420b, 420s, 420d, because these spectral lines quantized to non-zero value. In addition, the indexes of the frequencies of spectral lines a, 422b, s, 422d, 422h, 422i, 422j and 422k will not be in R, because these spectral lines are too close to the spectral lines 420, 420b, 420s and 420d. In contrast, the indexes of the frequencies of spectral lines e, 422f, 422g will be included in the set R, because these spectral lines themselves are quantized to zero and are located far enough away from any adjacent non-zero spectral lines.

The algorithm 300 also includes a second portion 320 of the decoding self-noise, where the index values of the noise (the"index" part of the code 320) is converted into decoded the value of noise ("nf_decoded" in code 300).

Code 300 also includes a third portion 330 homoserine identified spectral lines, i.e. spectral lines with frequency indexes i that belong in the set R. For this spectral values identified spectral lines (designated for example as x(i), where alternating current is i consistently takes values of all indexes of the frequencies of spectral lines, included in the set R) are the values of homoserine. Values homoserine obtained, for example, by multiplying the values of the decoded homoserine (nf_decoded) on a random number or a pseudo-random number (denoted as "random (-1, +1)), where random or pseudo-random number can take, for example, randomly or pseudo-random values -1 and +1. However, different software random or pseudo-random noise naturally possible.

Sumusuporta also shown in Figure 4. As can be seen in figure 4, the zero spectral values of the spectral lines e, 422f and 422g are replaced with the values of homoserine values (shown by the dotted lines in figure 4).

The evaluator parameter homoserine figure 5 and 6

Figure 5 shows the block diagram of the calculator option homoserine 500. The evaluator parameter homoserine is designed to produce a quantized spectral representation 510 of the audio signal and creation on its basis of the parameter homoserine 512. The evaluator parameter homoserine 500 includes an identifier of the spectral region 520, which is designed to produce a quantized spectral representation 510 of the audio signal and determine the spectral regions (e.g., spectral lines) of the quantized spectral representation 510, separated from the Nene is left spectral regions of the quantized spectral representation 510, at least one intermediate spectral region (e.g., spectral pattern)for information 522, describing the identified spectral region (for example, identified spectral lines). The evaluator parameter homoserine 500 also includes a transmitter jitter values 530, designed to retrieve information about the error quantization 532 and parameter definition of homoserine 512. For this purpose, the transmitter noise values are designed to selectively consider quantization errors in the identified spectral regions described information 522 to calculate the parameter homoserine 512.

Information about the error quantization 532 may be, for example, is identical to the energy information (or information about the intensity), describing the energy (or intensity) of those spectral lines quantized to zero quantized spectral representation 510.

The evaluator parameter homoserine 500 may further comprise a quantizer 540, which is designed to receive aquantance spectral representation 542 audio signal and determining a quantized spectral representation 510 of the audio signal. The quantizer 540 may have an adjustable resolution, which can, for example, be individually adjusted for each of the th spectral line, or spectral band (for example, depending on the psychoacoustic significance of spectral lines or spectral bands, which is revealed by using a psychoacoustic model). The functionality of the quantizer with variable resolution may be equal, the functionality described in the International Standards ISO/IEC 13818-7 and ISO/IEC 14496-3. In particular, the quantizer 540 may be configured so that the quantized spectral representation 510 of the audio signal will be spectral gaps or spectral holes, i.e. adjacent region adjacent spectral lines quantized to zero.

Moreover, your unquantized spectral representation 542 may serve as the information about the quantization errors 532, or information about the quantization errors 532 can be deduced from aquantance spectral representation 542.

Next will be described the functionality of the calculation parameter homoserine, which may be performed by the evaluator parameter homoserine 500. When calculating the parameter sumusuporta on the part of the coder, sumusuporta preferable to apply in the field quantization. In this way, the introduced noise is subsequently formed the inverse filter psychoacoustic significance.

The energy of the noise introduced by the decoder, is calculated and encoded on the side of the encoder on the trail of the relevant stages

1. To obtain quantized frequency lines.

2. To select only part of the spectrum.

3. To determine the spectral regions in the selected portion of the spectrum, where the length of the series of zeros is higher than the minimum run length.

4. To calculate the geometric mean of the quantization error from the previously discovered fields.

5. To uniformly quantize the geometric mean of 3 bits.

As regards the first stage, the quantized frequency lines can be obtained using the quantizer 540. Quantized frequency lines, therefore, can be represented in a quantized spectral representation 510.

Regarding the second stage, which can be considered optional, it should be noted that the calculation of homoserine it is preferable to carry out on the basis of the high-frequency part of the spectrum. In a preferred embodiment, the energy of the noise (called self-noise), is calculated only for the second half of the spectra, i.e. for high frequencies (but not for lower frequencies). Indeed, as a rule, high frequency (upper spectrum) are less important for perception than low frequencies, and zero-quantized values occur mainly in the second half of the spectra. Furthermore, the addition of noise at high frequencies are less likely to lead to a final noisy vosstanovlenieplastika sound.

Regarding the third stage, by limiting homoserine in spectral regions where there is a long series of zero-quantized values, you can avoid sumusuporta influenced niobocene the values too much. Thus, sumusuporta does not apply in the neighborhood by proximity to neopolen values, and the original tone of these lines then it is better preserved. The minimum length of the series is fixed to 8 in the preferred embodiment. This means that 8 lines surrounding niobocene values are not affected by homoserine (and therefore not taken into account in the calculation of the noise value).

As for the fourth stage, quantization errors in the quantized region are in the range of [minus 0.5, 0,5], and are considered to be uniformly distributed. The energy of the quantization error in the detected area is medium in the logarithmic region (i.e. the geometric mean). Own noise, nf, is then calculated as follows:

nf=power(10, sum (log10(E(x(i))))/(2*n)).

nf=power (10, sum (log10(E (x(i))))/(2*n)).

In the above formula, the sum () is the sum of the logarithmic energy log10 (E()), from individual lines x (i) within the detected regions, and n is the number of lines within these areas. Own noise, nf, is between 0 and 0.5. That is why the calculation allows to take into account the source spectral flatness reset values and then to get information about the characteristics of their tone / noise.

If the reset value is tonal, noise (computer apparatus 500) will go to zero, and low self-noise is added to the decoder (for example, the decoder 100, 200 described above). If zero values are really noisy, self-noise level will be high, and sumusuporta can be considered as highly parametric coding zeroed spectral lines, as PNS (Perceptual Noise Substitution) (replacing the perception of noise) (see also [4]).

Concerning the fifth stage, the quantization index ("index") from the self-noise is calculated as follows:

index=max(0, min(7, int(8-16*nf))).

The index is passed, for example, 3 bits.

Next will be described an algorithm for computing the parameter homoserine with reference to Fig.6, which shows the program pseudo-code 600 for this algorithm to obtain parameter homoserine in accordance with a variant of the invention. The algorithm 600 includes a first part - block-node 610 to identify areas that should be considered for the calculation of the parameter homoserine. Identified identified areas (e.g., spectral lines) are described by the set R, which may, for example, to enable the indexes of the frequencies of spectral lines ("index line") identified by SPECT the social lines. You can identify the spectral lines, which are quantized to zero and are located far enough from any other spectral lines quantized to non-zero value.

The first portion 610 of the program 600 may be identical to the first part 310 of the program 300. Accordingly, the quantized spectral representation ("quantized (x(i))")used in the algorithm 600 may be, for example, identical quantized spectral representation ("quantized x(i))")used in the algorithm 300 on the side of the decoder. In other words, the quantized spectral representation used on the side of the encoder can be transmitted in encoded form to the decoder in the transmission system that includes an encoder and a decoder.

The algorithm 600 includes a second portion 620 of the calculation of the self-noise. When calculating the self-noise are taken into consideration only those spectral region (or spectral lines)that are included in the set R, calculated in the first part 610 of the algorithm 600. As can be seen, the value of homoserine nf initially set to zero. The number of the considered spectral lines (n) is also initially set to zero. Next, the energy of all the spectral lines, the index lines, which are included in the set R are summed, and the energy of the spectral lines logarithmically before summation. For example, the logarithm on the basis of aniu 10 (log 10) from energy (E (x(i))) spectral lines can be accumulated. It should be noted that the real energy of spectral lines to quantization (labeled "E or energy (x(i))") added in logarithmic form. Also take into account the number of the considered spectral lines. Thus, after execution of the second portion 620 of the algorithm 600 variable nf indicates the logarithmic sum of the energies of the identified spectral lines to quantization, and variable n describes the number of identified spectral lines.

The algorithm 600 also includes a third portion 630 of quantization values nf, i.e. the logarithmic sum of the identified spectral lines. Displays the equation as described above or as shown in Fig.6, can be used.

Method of use according to Fig.7.

Fig.7 shows the block diagram of the method for obtaining samozabvennoi spectral representation of the audio signal based on the input spectral representation of the audio signal. Method 700 figure 7 includes the step 710 identify spectral regions of the input spectral representation of the audio signal separated from non-zero spectral regions of the input spectral representation, at least one intermediate spectral region, for receiving the identified spectral regions. Method 700 also includes the step 720, the choice is knogo the introduction of noise in the identified spectral region to obtain samozabvennoi spectral representation of the audio signal.

Method 700 can be complemented by any of the properties or functionality described herein in connection with the invented sumsemanntill.

Method of use in accordance with Fig

On Fig shows a block diagram of a method of obtaining parameter sumusuporta on the basis of the quantized spectral representation of the audio signal. The method 800 includes a step 810 identify spectral regions of the quantized spectral representation of the audio signal separated from non-zero spectral regions of the quantized spectral representation, at least one intermediate spectral region, to obtain identifizierung spectral regions. Method 800 also includes a step 820 selective consideration of quantization errors in the identified spectral regions are used to calculate parameter homoserine.

Method 800 can be complemented by any of the properties and functionality described herein in connection with the evaluator parameter homoserine.

A representation of the sound signal in Fig.9

Figure 9 shows a graphical representation of the representation of the audio signal in accordance with a variant of the invention. A representation of the sound signal 900 may, for example, serve as the basis for the input spectral representation 110. A representation of the sound signal 900 mo is et to take on the functionality of the encoded representation of the audio signal 212. A representation of the sound signal 900 can be obtained using the solver option homoserine 500, in which a representation of the sound signal 900 may, for example, include a quantized spectral representation 510 of the audio signal and the parameter homoserine 512, for example, in coded form.

In other words, the encoded representation of the audio signal 900 may represent the audio signal. The representation of the encoded audio signal 900 includes encoded quantized spectral representation of the audio signal and the encoded parameter homoserine. Option homoserine represents the quantization errors of spectral bands in the spectral representation of the domain, quantized to zero and separated from the spectral representation of the spectral domain, quantized to non-zero value, at least one intermediate spectral region.

Naturally, the representation of the sound signal 900 may be supplemented by any information set forth above.

Alternative ways to use

Depending on the specific requirements for the use of the invention can be implemented in hardware or in software. The invention can be applied using any digital media such as a disk is s, DVD, CD, ROM, FROM, EPROM, EEPROM, or flash cards, having established electronically-readable control signals that are compatible (or can work together with a programmable computer System, which is the appropriate method.

Some possible uses of the invention include the use of a data carrier having electronically-readable control signals, which are able to operate with a programmable computer system in which one of the methods described here.

As a rule, the use of the present invention may be implemented as a program product with program code, when this code is used to perform one of the methods when the program product runs on a computer.

Program code may be, for example, stored on electronic media.

Other embodiments of the invention include the use of a computer program for performing one of the methods described above, which is stored on a readable medium.

Thus, implementation of the invention is a computer program having a program code for performing one of the methods described herein when the computer program runs on a computer.

Another method of the invention, the methods used includes the e media (or digital storage device), containing a computer program for performing one of the methods described in this document.

Another form of the invention, the method is a data stream or a sequence of signals representing the computer program for performing one of the methods described in this document. The data stream or a sequence of signals may be, for example, is configured for transmission through the data transmission channel, for example via the Internet.

Another use case of the invention includes means for processing data, such as a computer or programmable logic device that is configured or adapted to perform one of the methods described in this document.

Another option is the use of the invention involves the use of a computer running a computer program for performing one of the methods described in this document.

In some ways the implementation of the invention, a programmable logic device (e.g., logical matrix, user-programmable) can be used to perform some or all of the functionality described in this document.

In some implementations, a logical matrix, user-programmable, can be used together with Mick what processora to execute one of the methods described in this document.

Conclusion

Summarizing the above, it should be noted that this invention improves such instrument sound encoding as "sumusuporta", because it takes into account the characteristics of the input signal and the decoded signal when the calculation of the parameters of homoserine on the side of the encoder and the use of noise on the side of the decoder.

Implementation of the invention assumes that the pitch of zero-quantized spectral lines is estimated and used to assess self-noise. This noise is then passed to the decoder, which performs sumusuporta zero quantized values that occur in separate regions of the spectra. These areas are chosen depending on the characteristics of the decoded spectra.

As for the context of the invention, it can be noted that the invention is associated with a transform-based coding, which uses scalar quantization on MDCT. MDCT coefficients of the pre-normalized curve, calculated on the perceptual importance. The curve is derived from the previous LPC (coding linear prediction Linear Prediction Coding) analysis by weighing LPC coefficients, as is done in the mode of TLC from the AMR-WB+(see [1]). Based on the weighted coefficients of the designed filter weigh-in & m is of perceptions, which is applied to the MDCT. The inverse filter is applied also on the side of the decoder after inverse MDCT. This inverse filter weighing perceptions shapes the quantization noise so that it minimizes or hides the perceived noise.

In all forms of the invention overcome the disadvantages of previous devices. Sumusuporta traditionally applied on a systematic basis for the zero-quantized values, considering only the spectral threshold envelope, the masking threshold or the threshold energy. Previous methods did not take into account any characteristics of the input signal, characteristics of the decoded signal. Thus, the traditional tools may introduce additional unwanted distortion, especially distortion noise, and to nullify the advantages of such a tool.

In contrast, the implementation of the present invention can improve sumusuporta limiting the undesirable distortion, as mentioned above.

Literature

[1] "Extended Adaptive Multi-Rate - Wideband (AMR-WB+) codec", 3GPP TS 26.290 V6.3.0, 2005-06 Technical Specification.

[2] Ragot et al, "ITU-T G.729.1: AN 8-32 Kbit/S Scalable Coder Interoperable with G.729 for Wideband Telephony and Voice Over IP", Vol.4, ICASSP 07, 15-20 April 2007.

[3] "AUDIO CODING", International Application No.: PCT/IB2002/001388, Applicant: KONINKLIJKE PHILIPS ELECTRONICS N.V. [NL/NL]; Groenewoudseweg 1 NL-5621 BA Eindhoven (NL). Inventors: TAORI, Rakesh; Prof Holstlaan 6, NL-5656 AA Eindhoven (NL) and VAN DE PAR, Steven, L., J. D., E.; Prof. Holslaan 6 NL-5656 AA Eindhoven (NL).

[4] Generic Coding of Moving Pictures and Associated Audio: Advanced Audio Coding. International Standard 13818-7, ISO/IEC JTC1/SC29/WG11 Moving Pictures Expert Group, 1997.

1. Homosapien (100) for receiving samozabvennoi spectral representation (112) of the audio signal based on the input spectral representation (110) of the audio signal, including:
ID spectral region (120), designed to identify spectral regions (a, 422f, 422g) of the input spectral representation (110), which are quantized to zero and separated from non-zero spectral regions (420, 420b, 420s, 420d) of the input spectral representation (110), at least one intermediate spectral region (a, 422b, s, 422d, 422h, 422i, 422j, 422k) to obtain the identified spectral regions (a, 422f, 422g); and
device for inserting noise (130), designed to selectively introduce noise in the identified spectral region (e, 422f, 422g) to obtain samozabvennoi spectral representation (112) of the audio signal.

2. Homosapien (100) according to claim 1, in which the identifier of the spectral region (120) is designed to identify, in accordance with the identified spectral regions, spectral lines (e, 422f, 422g) of the input spectral representation (110), which are quantized to zero and include at least the first specified number (4) low-frequency neighbour of their spectral lines (a, 422b, s, 422d; 422b, s, 422d, e, s, 422d, e, 422f), quantized to zero, and at least a second specified number (4) high-frequency neighboring spectral lines (422f, 422g, 422h, 422i, 422g, 422h, 422i, 422j; 422h, 422i, 422j, 422k), quantized to zero, in accordance with the identified spectral regions;
in which the first specified number (4) greater than or equal to 1, and the second specified number (4) greater than or equal to 1; and
in which the device for inserting noise (130) is designed to selectively introduce noise in the identified spectral lines (e, 422f, 422g), leaving a spectral line (420, 420b, 420s, 420d), quantized to non-zero value, and the spectral line (a, 422b, s, 422d, 422h, 422i, 422j, 422k), quantized to zero, but not having first given number (4) lower-frequency neighboring spectral lines quantized to zero, or the second given number (4) higher frequency neighboring spectral lines quantized to zero, is not subject to samozabvennuyu.

3. Homosapien (100) according to claim 2, in which the first specified number (4) is equal to the second specified number (4).

4. Homosapien (100) according to claim 1, in which homosapien is designed to make noise only in the spectral region of the upper part of the input spectral representation (110) of a sound signal, leaving the lower part of the input spectral representation (110) sounds the signal is not subject to samozabvennuyu.

5. Homosapien (100) according to claim 1, in which the identifier of the spectral region (120) is designed to summarize the quantized intensity values (quantized (x (i))) spectral regions in a given bilateral spectral range of this spectral region (i) to obtain the sum of the values (E) and estimate the amount of values (E)to determine whether this spectral region (i) identified spectral region or not.

6. Homosapien (100) according to claim 1, in which the identifier of the spectral region (120) is designed to scan a range of spectral regions of the input spectral representation (110) for the detection of related sequences (a - 422i; 422b - 422j; s - 422k) spectral regions, quantized to zero, and determining one or more Central spectral regions (a, 422f, 422g) detected a contiguous sequence as identified spectral regions.

7. The evaluator parameter homoserine (500) to determine the parameter homoserine (512) based on the quantized spectral representation (510) audio signal includes:
ID spectral region (520)designed to identify spectral regions (a, 422f, 422g) of the quantized spectral representation (510), separated from non-zero spectral the regions (420, 420b, 420s, 420d) of the quantized spectral representation (510), at least one intermediate spectral region (a, 422b, s, 422d, 422h, 422i, 422j, 422k) to obtain the identified spectral regions (a, 422f, 422g); and
the evaluator values of noise (530)designed to selectively consider quantization errors (power (x (i))) identified spectral regions (i) to calculate the parameter homoserine (512, nf).

8. The evaluator parameter homoserine (500) according to claim 7, in which the identifier of spectral bands (520) is used to identify spectral lines (e, 422f, 422g) of the input spectral representation (510), which are quantized to zero and which contain at least the first specified number (4) lower-frequency neighboring spectral lines (a, 422b, s, 422d; 422b, s, 422d, e, s, 422d, e, 422f), quantized to zero, and at least a second specified number (4) high-frequency neighboring spectral lines (422f, 422g, 422h, 422i, 422g, 422h, 422i, 422j; 422h, 422i, 422j, 422k), quantized to zero, in accordance with the identified spectral regions,
in which the first specified number (4) greater than or equal to 1, and the second specified number (4) greater than or equal to 1; and
in which the transmitter jitter values (520) is designed to selectively consider quantization errors identified range of the selected regions (i) to calculate the parameter homoserine, leaving in the calculation of the parameter without attention spectral line (420, 420b, 420s, 420d), quantized to non-zero value, and the spectral line (a, 422b, s, 422d, 422h, 422i, 422j, 422k), quantized to zero, but not having the first specified number (4) lower-frequency neighboring spectral lines quantized to zero, or the second specified number (4) high-frequency neighboring spectral lines quantized to zero.

9. The evaluator parameter homoserine (500) according to claim 7, in which the transmitter values of noise (530) is designed to take into account the actual energy of (x (i))) quantization errors in the identified spectral regions (i) to calculate the parameter homoserine (512, nf, nf_index).

10. The evaluator parameter homoserine (500) according to claim 7, in which the transmitter values of noise (530) is designed to determine the energy of the error naturalnego quantization energy (x (i))), distributed according to the multitude of identified spectral regions, compared with the energy of the error tone quantization, concentrated in a single spectral region or in many contiguous spectral lines.

11. The evaluator parameter homoserine (500) according to claim 7, in which the transmitter values of noise (530) is designed to calculate the sum of the logarithmic energy of the quantization error (log10 (energy (x (i)))) in the identified spectra the General areas of (i) to obtain parameter homoserine (512, nf, nf_index).

12. The method (700) for creating samozabvennoi spectral audio signal based on the input spectral representation of the audio signal, including:
identifying (710) spectral regions of the input spectral representation, separated from non-zero spectral regions of the input spectral representation, at least one intermediate spectral region to obtain the identified spectral regions; and
custom insert (720) noise in the identified spectral region to obtain samozabvennoi spectral representation of the audio signal.

13. The method (800) for calculating the parameter homoserine parameters on the basis of the quantized spectral representation of the audio signal includes:
identification (810) spectral regions of the quantized spectral representation, separated from non-zero spectral regions of the quantized spectral representation, at least one intermediate spectral region to locate the identified spectral regions; and
selective review (820) quantization errors in the identified spectral regions are used to calculate parameter homoserine.

14. The computer-readable storage medium recorded thereon a computer program for implementing the JV is soba indicated in paragraph 12, when a computer program running on the computer.

15. The computer-readable storage medium recorded thereon a computer program for implementing the method according to item 13, when the computer program runs on a computer.