JP2008058727A - Speech coding device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech coding device performing band-expansion coding wherein noise is reduced for a speech signal which has no or a very small signal of low frequency. <P>SOLUTION: When a low-frequency signal ratio detection part 25 detects the speech signal which has no or a very small signal of low frequency in an expansion band coder part 21 of the speech coding device, a coding system selection part 24a makes an HFC (High Frequency Coding) coding part 29 code a speech signal of an expansion band. The HFC coding part 29 performs parameter coding and waveform coding according to tonality of the speech signal of the expansion band. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a speech encoding apparatus, and more particularly to encoding processing of a high frequency speech signal.

  It is known that a speech signal is encoded by a band extension encoding method, that is, a low-frequency signal that is a basic portion and a high-frequency signal that is an extension portion are encoded by different encoding methods. ing. In this method, since there is a correlation between the high frequency signal and the low frequency signal, the high frequency signal is encoded by encoding power, noise level, and tone signal component as this correlation information. It is. According to this encoding method, highly efficient audio signal encoding is possible.

  For example, when an audio signal sampled at a certain sampling frequency is encoded by an AAC (Advanced Audio Coding) method, an Sbiaal (registered trademark, SBR, Spectral Band Replication) method (hereinafter, referred to as an SBR method). It is known that encoding is performed in combination with encoding according to).

  When an audio signal is encoded by the AAC method and the SBR method, a low-frequency signal obtained by converting the audio signal into a signal sampled at, for example, half the sampling frequency is encoded by the AAC method. To do.

  On the other hand, high frequency signals other than the low frequency are encoded by the SBR method. That is, the signal is converted into a subband signal by QMF (Quadrature Mirror Filter) analysis processing. First, grid information serving as a delimiter of envelope information is generated based on the detection result of the transient signal. The power information for each divided subband sample area (segment) is encoded as envelope information. A signal that cannot be expressed by envelope information is encoded as additional information.

  Finally, the signal encoded by the AAC method and the signal encoded by the SBR method are arranged in a predetermined format, and the encoding ends (see, for example, Patent Document 1).

Each sample signal encoded by the SBR method is based on a signal obtained by copying from a low-frequency subband signal obtained by decoding information encoded by the AAC method at the decoder. Restored.
JP-T-2005-510772 (page 2-6, FIG. 9)

  However, in the method disclosed in Patent Document 1 described above, when the low-frequency signal does not exist or is very small, the high-frequency signal cannot be accurately expressed and encoded. There is a problem that noise may be perceived when an audio signal is decoded.

  The present invention has been made to solve the above-described problems, and provides a speech coding apparatus that performs band extension coding with reduced noise on a speech signal in which a low-frequency signal does not exist or is very small. With the goal.

  In order to achieve the above object, a speech encoding apparatus according to the present invention encodes a speech signal by encoding a low frequency signal and a high frequency signal of the speech signal using different methods. A low-frequency signal encoding means for encoding the low-frequency signal, and a high-frequency signal for encoding the high-frequency signal by encoding correlation information of the high-frequency signal with the low-frequency signal. The high frequency signal encoding means calculates a ratio obtained by dividing the power of the high frequency signal by the power of the low frequency signal, and the calculated ratio is equal to or greater than a predetermined threshold value. The high frequency signal is encoded by multiplying the high frequency signal by a gain that is a monotonically decreasing function with respect to the ratio.

  The speech coding apparatus according to the present invention is a speech coding apparatus that encodes the speech signal by encoding the low frequency signal and the high frequency signal of the speech signal using different methods. A low-frequency signal encoding means for encoding a high-frequency signal, a ratio obtained by dividing the power of the high-frequency signal by the power of the low-frequency signal, and when the calculated ratio is equal to or less than a predetermined threshold, When the high frequency signal is encoded by encoding correlation information of the low frequency signal to the low frequency signal, and the calculated ratio exceeds a predetermined threshold, the correlation is not related to the low frequency signal. And high frequency signal encoding means for encoding the high frequency signal.

  ADVANTAGE OF THE INVENTION According to this invention, the audio | voice encoding apparatus which carries out the band expansion encoding of the audio | voice signal which a low-frequency signal does not exist or is very small by which noise was reduced can be provided.

  Embodiments of a speech encoding apparatus according to the present invention will be described below with reference to the drawings.

(First embodiment)
FIG. 1 is a block diagram showing a configuration of a speech encoding apparatus according to the first embodiment of the present invention. This speech encoding device is a device that receives and encodes a PCM speech signal 11 and generates an encoded speech signal 12 that is a coded signal, and encodes a high-frequency signal in the PCM speech signal 11. An extension band encoder unit 21, a low sampling frequency unit 31, an AAC encoder unit 32 that encodes a low-frequency signal in the PCM audio signal 11, and a stream formatter unit 41.

  FIG. 2 is a block diagram showing a configuration of the extension band encoder unit 21. The extension band encoder unit 21 includes a QMF analysis unit 22 that receives the PCM audio signal 11, a grid information generation unit 23, a gain control unit 24, a low frequency signal ratio detection unit 25, an additional parameter calculation unit 26, an envelope The information calculation unit 27 and the extension band stream formatter unit 28 connected to the stream formatter unit 41 are included.

  FIG. 3 is a block diagram illustrating a configuration of the low frequency signal ratio detection unit 25. The low frequency signal ratio detection unit 25 is connected to the low frequency band power calculation unit 25 a connected to the QMF analysis unit 22, the high frequency band power calculation unit 25 b connected to the QMF analysis unit 22, and the gain control unit 24. Power ratio calculation unit 25c.

  The operation of each unit of the speech encoding apparatus according to the first embodiment of the present invention configured as described above will be described with reference to FIGS.

  The QMF analysis unit 22 converts the received PCM audio signal 11 into the frequency domain, and generates a subband signal. The grid information generation unit 23 generates grid information serving as a delimiter of envelope information based on the detection result of the transient signal of the subband signal generated by the QMF analysis unit 22.

  The low frequency signal ratio detection unit 25 receives the subband signal generated by the QMF analysis unit 22, and detects the ratio of the power of the low frequency signal to the power of the high frequency signal. That is, a ratio is generated by dividing the power (Phigh) of the high frequency signal by the power (Plow) of the low frequency signal.

  The gain control unit 24 detects the power (Plow) of the low frequency signal encoded by the AAC encoder unit 32 detected by the low frequency signal ratio detection unit 25 and the high frequency signal encoded by the extension band encoder unit 21. Based on the ratio (High / Plow) to the signal power (High), a signal is generated by giving a predetermined gain to the high-frequency subband signal generated by the QMF analyzer 22.

  Here, the gain is 1 if the ratio is equal to or less than a predetermined value. And if the ratio exceeds a predetermined value, it is a monotonically decreasing function that takes a value of less than 1. That is, if the power of the high frequency signal is constant, the smaller the power of the low frequency signal, the smaller the gain, in other words, the greater the attenuation.

  With reference to FIG. 3, the operation of each unit constituting the low-frequency signal ratio detection unit 25 will be described. The low band power calculation unit 25a calculates the power (Plow) of the low band subband signal of the subband signal generated by the QMF analysis unit 22. The high band power calculator 25b calculates the power (Ph) of the high band sub-band signal of the sub-band signal generated by the QMF analyzer 22. Then, the power ratio calculation unit 25c uses the power (Phigh) of the high frequency signal calculated by the high frequency band power calculation unit 25b as the power (Plow) of the low frequency signal calculated by the low frequency band power calculation unit 25a. The power ratio (High / Plow) is generated by obtaining the quotient by dividing by.

  When this ratio is generated, the position where the low frequency signal and the high frequency signal are separated in the frequency domain, that is, the position where the AAC coding and the extended band coding are separated depends on the sampling frequency and the coding rate. To do. Therefore, the power ratio calculation unit 25c performs correction by multiplying the ratio (Nlow / Nhigh) of the low frequency subband number Nlow and the high frequency subband number Nhigh. As a result, the corrected power ratio Pr becomes (Phigh / Plow) × (Nlow / Nhigh).

  Referring back to FIG. 2, the description returns to the operation of each unit constituting the extension band encoder unit 21. The additional parameter calculation unit 26 detects signals that are subbands divided by the grid information generated by the grid information generation unit 23 and cannot be expressed by the envelope information from the signals transmitted by the gain control unit 24. . Then, additional information such as a noise level expressing the signal is obtained, and a parameter indicating the additional information is generated.

  The envelope information calculation unit 27 encodes power information for each region (segment) of the high-frequency subband sample as envelope information for the subband signal divided by the grid information generated by the grid information generation unit 23. To do.

  The extension band stream formatter unit 28 includes the grid information generated by the grid information generation unit 23, the parameter indicating the additional information generated by the additional parameter calculation unit 26, and the high frequency band encoded by the envelope information calculation unit 27. The audio signal is received, and the audio signal encoding information of a predetermined extension band, that is, a high frequency is arranged into a stream of a predetermined format and transmitted. The encoded high-frequency audio signal adjusted to a predetermined format by the extension band stream formatter unit 28 is transmitted to the stream formatter unit 41.

  The low sampling frequency unit 31 receives the PCM audio signal 11 and generates a low-frequency signal included in the signal by down-sampling. For example, a combination of normal AAC coding and SBR coding produces a signal sampled at a frequency that is half the frequency at which the input signal was sampled. The process to be generated is not limited to a half frequency. The AAC encoder unit 32 receives the low frequency audio signal generated by the low sampling frequency unit 31, encodes it by the AAC method, and transmits the encoded signal.

  The stream formatter unit 41 receives the encoded high frequency audio signal adjusted to a predetermined format by the extension band stream formatter unit 28 and the low frequency audio signal encoded by the AAC encoder unit 32. The high frequency audio signal and the low frequency audio signal are arranged into a stream of a predetermined format, and the encoded audio signal 12 which is an encoded audio signal is transmitted.

  As described above, the low-frequency signal ratio detection unit 25 performs encoding by giving a high attenuation to the high-frequency audio signal as the power of the low-frequency audio signal is smaller than that of the high-frequency audio signal. This is due to the following reason. If the power of the low frequency audio signal is small, the high frequency signal obtained by copying it is obtained by performing envelope correction (gain increase) based on the envelope information with low frequency resolution. This is because when the correlation of the band signals is low, distortion increases and is perceived as noise. Further, since the low-frequency signal important for human hearing is small and the masking effect for high-frequency distortion does not work, noise is easily perceived. Therefore, the above-described attenuation has an effect of making it difficult for the user who has listened to the signal obtained by decoding the encoded audio signal 12 to feel uncomfortable due to the noise.

(Second Embodiment)
The difference between the speech coding apparatus according to the second embodiment of the present invention and the speech coding apparatus according to the first embodiment resides in the extension band encoder unit 21. Therefore, the extension band encoder unit 21 according to the second embodiment will be described. FIG. 4 is a block diagram showing a configuration of the extension band encoder unit 21. In the extension band encoder unit 21 according to the second embodiment, the same parts as those of the extension band encoder unit 21 according to the first embodiment are denoted by the same reference numerals and description thereof is omitted.

  The extension band encoder unit 21 includes a QMF analysis unit 22 that receives the PCM audio signal 11, a grid information generation unit 23, an encoding scheme selection unit 24a, a low frequency signal ratio detection unit 25, and an additional parameter calculation unit 26. And an envelope information calculation unit 27, an extension band stream formatter unit 28a connected to the stream formatter unit 41, and an HFC (High Frequency Coding) encoding unit 29.

  That is, the extension band encoder unit 21 includes an encoding method selection unit 24a instead of the gain control unit 24, as compared with the extension band encoder unit 21 according to the first embodiment, and an extension band stream formatter unit 28. Instead, an extended band stream formatter unit 28a is provided, and an HFC encoding unit 29 is further provided.

  FIG. 5 is a block diagram showing a configuration of the HFC encoding unit 29. The HFC encoding unit 29 includes a tonality calculation unit 29a connected to the encoding method selection unit 24a and the grid information generation unit 23, an encoding mode selection unit 29b connected to the grid information generation unit 23, and an extension band stream The parameter encoding unit 29c is connected to the formatter unit 28a, and the waveform encoding unit 29d is connected to the extension band stream formatter unit 28a.

  The operation of the extension band encoder unit 21 configured as described above according to the second embodiment of the present invention will be described with reference to FIGS.

  The encoding scheme selection unit 24a receives the power ratio (Pr) generated by the low-frequency signal ratio detection unit 25, and if this ratio is equal to or less than a predetermined threshold, the high-frequency signal is converted into an additional parameter calculation unit. 26 and the envelope information calculation unit 27 are selected to encode, and the high-frequency subband signal generated by the QMF analysis unit 22 is sent to these processing units for encoding.

  On the other hand, when the power ratio (Pr) exceeds a predetermined threshold, the high-frequency signal is selected to be encoded by the HFC encoder 29 and the high-frequency subband signal generated by the QMF analyzer 22 is selected. The data is sent to the HFC encoder 29 for encoding.

  The tonality calculation unit 29a receives the high frequency sub-band signal generated by the QMF analysis unit 22, and calculates the tonality for each sub-band signal in the region delimited by the grid generated by the grid information generation unit 23. calculate. When calculating tonality, for example, a linear prediction gain is used. At this time, the value of the tonality is weighted according to the power so that the characteristics of the signal having higher power can be easily reflected. The weighting is performed by normalizing with the power value of the subband signal having the maximum power value within the scale factor band.

  The encoding mode selection unit 29b receives the tonality calculated by the tonality calculation unit 29a, and based on the tonality of each subband signal in the scale factor band of the region delimited by the grid generated by the grid information generation unit 23. Then, the encoding mode for each scale factor band is selected. That is, the determination is made based on whether the highest value of the tonality of each subband signal in the scale factor band is equal to or greater than a predetermined threshold value.

  When the tonality is equal to or less than a predetermined threshold, the scale factor band is determined to be a noise-like signal, and the encoding mode selection unit 29b includes parameter information including power information related to the signal in the region divided by the grid. Is encoded by the parameter encoding unit 29c. On the other hand, when the tonality exceeds a predetermined threshold, the waveform encoding unit 29d is encoded with the signal.

  The parameter encoding unit 29c encodes a flag and power information indicating that it is encoding of a noise-like signal for each segment. The waveform encoding unit 29d encodes each subband sample. The waveform coding is performed according to the bit rate of the target extension band stream, in other words, according to the amount of information transmitted by the target extension band stream formatter unit 28a. To decide. Then, Huffman coding is performed on the difference value of each quantized subband sample.

  The extension band stream formatter unit 28a includes the grid information generated by the grid information generation unit 23, the parameter indicating the additional information generated by the additional parameter calculation unit 26, and the high frequency band encoded by the envelope information calculation unit 27. In addition to the audio signal, the parameter information encoded by the parameter encoding unit 29c and the high frequency audio signal encoded by the waveform encoding unit 29d are received. Then, these received signals are arranged into a predetermined format stream in which a predetermined extension band, that is, a high frequency audio signal is encoded, and transmitted.

  FIG. 6 shows an example in which each subband is encoded by any method by the speech encoding apparatus according to the second embodiment. Here, it is assumed that it is divided into 16 subbands. The six sub-bands in the low band are AAC encoded by the AAC encoder unit 32. In FIG. 6, these subbands are hatched from upper right to lower left.

  The high-frequency subbands have a first scale factor band (sfb1) having three subbands, a second scale factor band (sfb2) having three subbands, and four subbands in order of decreasing frequency. It is divided into a third scale factor band (sfb3).

  Among the scale factor bands, the subband having the highest tonality is determined by the encoding mode selection unit 29b, and the tonality is compared with a threshold value for each scale factor band. In FIG. 6, the sub-band having the highest tonality among the scale factor bands is hatched from the upper left to the lower right.

  If the tonality is equal to or less than a predetermined threshold, the signal of the scale factor band is encoded by the parameter encoding unit 29c. On the other hand, when the tonality exceeds a predetermined threshold value, the waveform encoding unit 29d performs encoding.

  In the example shown in FIG. 6, since the maximum tonality at sfb1 exceeds the threshold TH1 for sfb1 (tonal (max)> TH1), the signal of sfb1 is encoded by the waveform encoding unit 29d. Further, since the maximum tonality in sfb2 exceeds the threshold TH2 for sfb2 (tonal (max)> TH2), the signal of sfb2 is encoded by the waveform encoding unit 29d.

  Since the maximum tonality in sfb3 is less than the threshold TH3 for sfb3 (tonal (max) <TH3), the signal of sfb3 is encoded as parameter information including power information by the parameter encoding unit 29c. Here, in consideration of auditory characteristics, the higher the frequency, the easier the parameter coding is selected, so that the threshold value is increased as the frequency is higher (TH1 <TH2 <TH3).

  When the operation of the extension band encoder unit 21 and the operation of the extension band encoder unit 21 according to the first embodiment are compared, when the power ratio (Pr) is high, the high frequency signal is distorted in terms of audibility. By performing waveform encoding by the waveform encoding unit 29d on a tonal signal that is easily perceived, a high-frequency signal can be encoded with low distortion.

  On the other hand, in that case, the number of bits of the code created by the extension band encoder unit 21 may be large. However, when the ratio is high, it means that there is no low frequency signal or is very small. As a result, the number of bits allocated to the low frequency signal generated by the AAC encoder unit 32 is small. As a result, it is possible to prevent the number of bits of the encoded audio signal 12 from becoming excessive.

(Other embodiments)
The speech encoding apparatus according to the embodiment of the present invention may be a computer that operates using a program. In addition, the present invention can naturally be applied to any device that encodes an audio signal. Here, it goes without saying that the audio signal includes a voice generated by a person, a musical sound, and any other sound. Moreover, you may combine suitably the element demonstrated in said embodiment. The present invention is not limited to the above configuration, and various modifications are possible.

1 is a block diagram showing a configuration of a speech encoding apparatus according to a first embodiment of the present invention. The block diagram which shows the structure of the extension zone | band encoder part which concerns on the 1st Embodiment of this invention. The block diagram which shows the structure of the low-pass signal ratio detection part which concerns on the 1st Embodiment of this invention. The block diagram which shows the structure of the expansion zone | band encoder part which concerns on the 2nd Embodiment of this invention. The block diagram which shows the structure of the HFC encoding part which concerns on the 2nd Embodiment of this invention. The figure which shows an example of selection of the encoding system of each subband which concerns on the 2nd Embodiment of this invention.

Explanation of symbols

11 PCM audio signal 12 Encoded audio signal 21 Extended band encoder unit 22 QMF analysis unit 23 Grid information generation unit 24 Gain control unit 24a Coding method selection unit 25 Low frequency signal ratio detection unit 25a Low frequency band power calculation unit 25b High frequency Band power calculation unit 25c Power ratio calculation unit 26 Additional parameter calculation unit 27 Envelope information calculation unit 28, 28a Extended band stream formatter unit 29 HFC encoding unit 29a Tonality calculation unit 29b Encoding mode selection unit 29c Parameter encoding unit 29d Waveform code Conversion unit 31 low sampling frequency conversion unit 32 AAC encoder unit 41 stream formatter unit

Claims (6)

An audio encoding device that encodes the audio signal by encoding the low-frequency signal and the high-frequency signal of the audio signal in different manners,
Low-frequency signal encoding means for encoding the low-frequency signal;
High frequency signal encoding means for encoding the high frequency signal by encoding correlation information of the high frequency signal to the low frequency signal;
The high frequency signal encoding means calculates a ratio obtained by dividing the power of the high frequency signal by the power of the low frequency signal, and when the calculated ratio is equal to or greater than a predetermined threshold, monotonously decreases with respect to the ratio. A speech coding apparatus, wherein the high frequency signal is multiplied by a gain as a function and coded. An audio encoding device that encodes the audio signal by encoding the low-frequency signal and the high-frequency signal of the audio signal in different manners,
Low-frequency signal encoding means for encoding the low-frequency signal;
Calculate a ratio obtained by dividing the power of the high-frequency signal by the power of the low-frequency signal, and if the calculated ratio is equal to or less than a predetermined threshold, encode correlation information of the high-frequency signal to the low-frequency signal Encoding the high frequency signal, and when the calculated ratio exceeds a predetermined threshold, high frequency signal encoding means for encoding the high frequency signal irrespective of the correlation with the low frequency signal; A speech encoding apparatus comprising: The high frequency signal encoding means, when the calculated ratio exceeds a predetermined threshold, divides the high frequency signal into a plurality of scale factor bands, calculates the tonality for each of the divided scale factor bands, If the calculated tonality is less than or equal to the tonality threshold, the signal included in the scale factor band is parameter-encoded. If the calculated tonality exceeds the tonality threshold, the signal included in the scale factor band is waveform encoded. The speech encoding apparatus according to claim 2, wherein: The speech encoding apparatus according to claim 3, wherein the tonality for each scale factor band is a maximum value of the tonality of a subband signal included in the scale factor band. The speech encoding apparatus according to claim 3, wherein the tonality threshold value is a larger value with respect to the higher scale factor band. The speech coding apparatus according to claim 3, wherein the parameter coding is coding of a parameter including power information of the signal.

Images