JPWO2009038115A1 - Speech coding apparatus, speech coding method, and program - Google Patents

Abstract

High-efficiency voice coding method reduces sound quality degradation when distributing music and melody to mobile terminals. The speech coding apparatus includes an auditory masking shaping processing unit 120 that suppresses and outputs a signal component that is audibly unnecessary due to the auditory masking effect, and outputs an output signal of the auditory masking shaping processing unit. A speech encoding processing unit 130 that executes speech encoding processing that performs speech compression encoding and outputs a bit stream (FIG. 1).

Description

[Description of related applications]
The present invention is based on the priority claim of Japanese patent application: Japanese Patent Application No. 2007-245547 (filed on Sep. 21, 2007), the entire description of which is incorporated herein by reference. Shall.
The present invention relates to a voice encoding device, a voice encoding method, and a program for improving the sound quality of music signals, melody signals, and the like transmitted by a voice encoding method.

  In recent years, services for distributing music and melody to mobile terminals have become widespread. For example, when a call is made to the other party using a mobile phone, a music signal is sent from the voice processing device prepared on the network side to the mobile phone as a ringback melody, or music content is sent from the voice processing device to the mobile phone. The service etc. delivered to are mentioned.

  When such a service is realized, a music signal or music content is preliminarily used by using the same method as the voice coding method (for example, the AMR coding method of Non-Patent Document 1) mounted on a portable terminal serving as a playback device. The data is distributed as a compressed and encoded bit stream.

  Although not intended for deterioration of sound quality when the music signal or music content is transmitted, Patent Document 1 is cited as a document that attempts to improve sound quality. In Patent Document 1, when the amplitude and phase of a plurality of encoded harmonics are input and decoded, and the decoded harmonics are harmonically masked by other harmonics, A speech decoding apparatus including an amplitude partial suppressing unit that suppresses the amplitude of harmonics is disclosed. The document does not disclose a configuration for encoding decoded speech.

  Japanese Patent Application Laid-Open No. 2004-26883 also includes a determination unit that determines whether or not the input sound is a non-speech signal, and a route selection unit that selects whether or not to pass the audibility correction filter based on the determination result. An encoding device and a speech decoding device are disclosed. The non-speech signal in this document refers to a data signal. When the input signal is non-speech (data signal), the non-speech signal does not pass through the audibility correction filter. The output is via (see paragraphs 0032 and 0099). Also, this document does not disclose a configuration for encoding decoded speech.

JP-A-6-332496 Japanese Patent Laid-Open No. 9-50298 3GPP TS 26.090 v. 3.1.0 "AMR Speech Codec Transcoding Functions", 1999 "Digital Coding of Waveforms", Prentice Hall, 1990 (DIGITAL CODING OF WAVEFORMS, PRINCIPLES AND APPLICATIONS TO SPEECH AND VIDEO, PENTICE-HALL, 1990.) "Multirate Systems and Filterbanks", Prentice Hall, 1993 (MULTIIRATE SYSTEMS AND FILTER BANKS, PENTICE-HALL, 1993.) “Psychoa Caustics”, Springer, 1999 (PSYCHOACUSTICS, SPRINGER, 1999.) "IEE International Conference on Acoustic Speech and Signal Processing, 25.1.1, 937-940, March 1985 (IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, 25.1.1, MAR, 1985, pp. 937-940)

  The entire disclosures of Patent Documents 1 and 2 and Non-Patent Documents 1 to 5 are incorporated herein by reference. The following is an analysis of the related art according to the present invention.

  A CELP (Code Excitation Linear Prediction) type audio encoding method such as the above AMR encoding method is optimized in principle for call speech, and there is little degradation in sound quality even if the audio signal is compression encoded. When the music signal is compression-encoded, the sound quality is greatly deteriorated. For this reason, when a melody or music content is distributed by these voice encoding methods, there is a problem that the sound quality is greatly deteriorated during reproduction on a portable terminal.

  This is thought to be because, in a speech coding system optimized for speech signals, music signal components that cannot be modeled become noise due to compression coding and are superimposed on the playback signal, and this noise is heard. .

  The present invention has been made in view of the above-described problems, and is a sound quality when distributing music and melody to a mobile terminal that needs to distribute a bitstream compressed and encoded by an audio encoding method. It is an object to provide a speech coding apparatus, speech coding method, and program capable of reducing deterioration of the sound.

  According to a first aspect of the present invention, an auditory masking shaping processing unit that suppresses and outputs a signal component that is audibly unnecessary due to an auditory masking effect among signal components of an audio signal, and the auditory masking shaping processing unit A speech encoding device is provided that includes a speech encoding unit that performs speech encoding processing that outputs a bit stream by performing speech compression encoding of the output signal.

  According to the second aspect of the present invention, the speech coding apparatus suppresses and outputs a signal component that is audibly unnecessary due to the auditory masking effect among the signal components of the input speech signal, and the speech coding apparatus. However, there is provided a speech coding method for speech compression coding the shaped signal in which the auditory unnecessary signal component is suppressed and outputting a bit stream.

  According to a third aspect of the present invention, there is provided a program that is executed by a computer that constitutes a speech coding apparatus, and suppresses a signal component that is audibly unnecessary due to an auditory masking effect among signal components of an input speech signal. A program for causing the computer to execute an audio masking shaping process to be output and an audio encoding process to output a bit stream by audio compression encoding the shaped signal subjected to the audio masking shaping process is provided.

  ADVANTAGE OF THE INVENTION According to this invention, deterioration of the sound quality at the time of delivering a music and a melody can be reduced with respect to the portable terminal which needs to deliver the bit stream compression-encoded by the audio | voice coding system. The reason is that a configuration is adopted in which components that are audibly unnecessary and components that cause deterioration are removed in advance.

It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the 1st Embodiment of this invention. It is a block diagram showing the example of a structure of the auditory masking shaping process part of FIG. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the 2nd Embodiment of this invention.

Explanation of symbols

100, 140 Terminal 120 Auditory masking shaping processing unit 122 Frequency conversion unit 124 Smoothing unit 126 Shaping unit 128 Frequency inverse conversion unit 130 Speech coding processing unit 250_1, 250_2 switching unit

  Of the signal components of the audio signal, a means for suppressing and outputting a signal component that is audibly unnecessary by the auditory masking effect, and a bit by compressing and encoding the output signal in which the signal component that is audibly unnecessary is suppressed A speech encoding device including means for outputting a stream can be developed in the following form.

  The process of suppressing and outputting the auditory unnecessary signal component is unnecessary auditoryly due to the presence of a high-level signal component (masker) on the frequency axis for each predetermined time interval for the decoded signal. This can be realized by removing the frequency component (masky) to be output and returning it to the time axis.

  The processing means that suppresses and outputs the signal components that are audibly unnecessary includes, for example, a frequency conversion unit that converts the frequency of a block composed of input audio signals, and smoothing that smoothes the output signal of the frequency conversion unit A shaping unit that removes unnecessary frequency components in the output signal of the frequency conversion unit using the output signal of the smoothing unit as a masking threshold, and the output signal of the shaping unit is inversely transformed and shaped And a frequency inverse transform unit that outputs the received signal.

  Instead of using the masking threshold to remove unnecessary frequency components in the output signal of the frequency converter, or in combination with this method, a level of a predetermined number of frequency components on the frequency axis remains. A method of removing low frequency components can be used.

  Further, a frequency component in a predetermined band can be the removal target.

  The speech encoding apparatus may further include a switching unit that analyzes characteristics of the input speech signal and switches whether to perform output via the auditory masking shaping processing unit. The switching unit may be configured to output after suppressing the auditory unnecessary signal component when the input audio signal has a characteristic of a music signal.

  Next, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

[First Embodiment]
FIG. 1 is a diagram showing the configuration of a speech encoding apparatus according to the first embodiment of the present invention. Referring to FIG. 1, the speech encoding apparatus includes an auditory masking shaping processing unit 120 and a speech encoding processing unit 130. The auditory masking shaping processing unit 120 and the speech coding processing unit 130 described above can be realized not only by a circuit configuration but also by a program that causes a computer to function as each processing unit described above.

  The auditory masking shaping processing unit 120 performs processing based on auditory psychological analysis on the frequency axis for the audio signal input from the terminal 100, suppresses components that are determined not to affect hearing, and returns them to the time axis. Output.

  The speech coding processing unit 130 receives the output signal of the auditory masking shaping processing unit 120, divides the signal at predetermined time intervals, performs speech coding processing, and outputs a compression-coded bit stream through the terminal 140. To do. For speech coding, for example, AMR speech coding described in Non-Patent Document 1 can be used, and in this case, the interval between the output signals described above is 20 ms. Here, it is assumed that the entire description of Non-Patent Document 1 is incorporated by reference in this document.

  Next, a detailed configuration of the auditory masking shaping processing unit 120 in FIG. 1 will be described with reference to FIG.

  Referring to FIG. 2, the auditory masking shaping processing unit 120 according to the present embodiment includes a frequency conversion unit 122, a smoothing unit 124, a shaping unit 126, and a frequency inverse conversion unit 128.

  The frequency converter 122 converts the audio signal input from the terminal 100 in FIG. 1 into a component on the frequency axis, generates a converted signal, and outputs the converted signal to the smoothing unit 124 and the shaping unit 126.

  When generating the converted signal, the frequency converting unit 122 collects a plurality of input signal samples to form one block, and applies the frequency conversion to this block. Examples of frequency conversion include Fourier transform, cosine transform, and KL (Kalunen label) transform. Non-patent document 2 discloses a technique related to specific operations of these conversions. Here, it is assumed that the entire description of Non-Patent Document 2 is incorporated herein by reference.

  In addition, it is possible to employ a configuration in which the frequency converter 122 weights one block of input signal samples with a window function when generating the converted signal. As such window functions, window functions such as Hamming, Hanning (Han), Kaiser, and Blackman are known. A more complicated window function can also be used. Non-patent document 3 discloses a technique related to these window functions. Here, the entire description of Non-Patent Document 3 is incorporated herein by reference.

  Further, when the frequency converting unit 122 forms one block from the input signal samples, each block may be overlapped. For example, when an overlap of 50% of the block length is applied, the last (second half) 50% of the signal samples belonging to a block is set to the first (first half) 50% of the signal samples belonging to the next block. Are used in duplicate in a plurality of blocks. Non-patent document 3 discloses a technique related to blocking and conversion having overlap.

  Furthermore, the above-described frequency conversion unit 122 may be configured by a band division filter bank that includes a plurality of band pass filters and divides a received input signal into a plurality of frequency bands. . Each frequency band of the band division filter bank may be equally spaced or unequal. In the case of dividing the band at unequal intervals, the time resolution can be reduced by dividing the band into a narrow band in the low band, and the time resolution can be increased by dividing the band in a high band. Typical examples of unequal interval division include octave division in which the band is successively halved toward the low band and critical band division corresponding to human auditory characteristics. A technique related to the band division filter bank and its design method is disclosed in Non-Patent Document 3.

  The smoothing unit 124 smoothes the conversion signal input from the frequency conversion unit 122 described above, and outputs the smoothed conversion signal to the shaping unit 126. Examples of the smoothing method include a method using the auditory masking effect disclosed in Non-Patent Document 4. For example, a smoothed transformed signal can be generated by convolving the transformed signal on the frequency axis using a function in which a certain frequency component masks nearby frequency components. Here, the entire description of Non-Patent Document 4 is incorporated herein by reference.

  As a simple smoothing method, S2 (n) may be calculated by the following equation [Equation 1], and a signal obtained by lowering the energy level of S2 (n) may be used as a smoothed signal. Here, max (x, y) represents the larger of x and y. E (n) is the energy of the converted signal and N is the block size.

[Equation 1]
S1 (0) = E (0)
S1 (n) = max (E (n), a × S1 (n−1)) (n = 1,..., N−1)
S2 (N-1) = S1 (N-1)
S2 (n) = max (S1 (n), b × S2 (n + 1)) (n = N−2,..., 0)

  The smoothed conversion signal calculated in this way is obtained by smoothing the energy level of the original conversion signal, and can be used as a masking threshold. That is, a frequency component having an energy level lower than the masking threshold value is to be removed as it is not perceptually perceived.

  The shaping unit 126 shapes the converted signal using the smoothed converted signal input from the smoothing unit 124. More specifically, the shaping unit 126 shapes the converted signal by removing a frequency component having an energy level smaller than that of the smoothed converted signal.

  At this time, the shaping unit 126 shapes the converted signal by leaving only a predetermined number of frequency components and removing other frequency components in descending order of the energy level ratio of the converted signal to the smoothed converted signal. You may do it. Further, the shaping unit 126 may remove only the low frequency, only the high frequency, or both the low frequency and the high frequency as the band limitation.

  The frequency inverse transform unit 128 inversely transforms the shaped conversion signal to generate a shaped signal, and outputs the shaped signal as an output signal of the auditory masking shaping processing unit 120. As the inverse transformation executed in the frequency inverse transformation unit 128, it is desirable to select an inverse transformation corresponding to the transformation applied by the frequency transformation unit 122. For example, when the frequency conversion unit 122 collects a plurality of input signal samples to form one block and applies frequency conversion to this block, the frequency inverse conversion unit 128 applies the corresponding inverse to the same number of samples. Apply transformation. In addition, when the frequency converting unit 122 constitutes one block from a plurality of input signal samples, if the blocks are allowed to overlap (overlap), the frequency inverse converting unit 128 correspondingly performs the inverse conversion. The same overlap is applied to later signals. Further, when the frequency converting unit 122 is configured by a band division filter bank, the frequency inverse converting unit 128 is configured by a band synthesis filter bank. A technique related to the band synthesis filter bank and its design method is disclosed in Non-Patent Document 3.

  As described above, the smoothing unit 124 and the shaping unit 126 use auditory characteristics such as a masking effect to remove a signal component that is not perceptually recognized from the shaped signal generated in this way (not audibly unnecessary). After the component is removed), the signal is returned to the time axis.

  Therefore, the auditory masking shaping processing in the auditory masking shaping processing unit 120 is performed by CELP-type analysis / synthesis coding represented by an AMR coding method or the like (details are disclosed in Non-Patent Document 5. All descriptions of Non-Patent Document 5). The contents are assumed to be incorporated by reference in this document.) When used as preprocessing, the linear prediction coefficient and pitch period are analyzed by analyzing the shaped signal from which auditory unnecessary components have been removed. And the like, and the sound quality of the decoded signal is improved.

[Second Embodiment]
Subsequently, a second embodiment of the present invention in which a change is made to the first embodiment of the present invention will be described.

  FIG. 3 is a block diagram showing a configuration of a speech encoding apparatus according to the second embodiment of the present invention. In FIG. 3, the constituent elements having the same numbers as those in FIGS. 1 and 2 perform the same operations as those in FIGS.

  In FIG. 3, the switching unit 250_1 extracts various feature parameters by dividing the audio signal input from the terminal 100 at predetermined time intervals, and performs auditory masking shaping processing based on the obtained feature parameters. Determine if it is better. For example, if the switching unit 250_1 determines that the musicality is strong (has a characteristic of a music signal) as a result of the determination by combining the characteristic parameter values, the switching unit 250_1 is input from the terminal 100 to the auditory masking shaping processing unit 120. Audio signal is output.

  On the other hand, when it is determined that the voice property is strong (musicality is weak) as a result of the combination parameter value determination, the switching unit 250_1 outputs the voice signal input from the terminal 100 to the switching unit 250_2. . Here, the switching unit 250_2 performs a switching operation in synchronization with the switching unit 250_1.

  As described above, according to the present embodiment, it is possible to accurately capture a music signal and input the audio signal input from the terminal 100 to the auditory masking shaping processing unit 120. Can be further reduced. Further, according to the present embodiment, since it is not necessary to consider that a voice signal with strong voice characteristics is input to the auditory masking shaping processing unit 120, the processing in the auditory masking shaping processing unit 120 is more efficient. It becomes possible.

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments, and further modifications and replacements are possible without departing from the basic technical idea of the present invention. -Adjustments can be made.

The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments, and further modifications and replacements are possible without departing from the basic technical idea of the present invention. -Adjustments can be made.
[Appendix 1-Claim 11 at the time of international application]
The block composed of the input audio signal is frequency converted,
Smoothing the frequency converted signal;
Using the smoothed signal as a masking threshold, remove unnecessary frequency components from the frequency converted signal,
The speech encoding method according to claim 9 or 10, wherein the auditory unnecessary signal component is suppressed by inversely transforming the signal from which the unnecessary frequency component is removed.
[Appendix 2-Claim 12 at the time of international application]
The speech coding method according to any one of claims 9 to 11, wherein the audibly unnecessary signal component is suppressed by removing the frequency component so that a predetermined number of frequency components remain on the frequency axis.
[Appendix 3-Claim 13 at the time of international application]
The speech coding method according to any one of claims 9 to 12, wherein a signal component that is audibly unnecessary is suppressed by removing a frequency component of a predetermined band.
[Appendix 4-International application claim 14]
14. The auditory unnecessary signal component is suppressed after analyzing characteristics of an input audio signal and determining whether or not to suppress the auditory unnecessary signal component. The speech encoding method described in 1.
[Appendix 5-International filing claim 15]
15. The speech encoding method according to claim 14, wherein when the input speech signal has characteristics of a music signal, the auditory unnecessary signal component is suppressed.
[Appendix 6-Claim 16 at the time of international application]
A program to be executed by a computer constituting a speech encoding device,
Auditory masking shaping processing that suppresses and outputs signal components that are audibly unnecessary by the auditory masking effect among the signal components of the input audio signal;
A program that causes the computer to execute voice coding processing for voice compression coding the shaped signal subjected to the auditory masking shaping processing and outputting a bit stream.

Claims (16)

Auditory masking shaping processing unit that suppresses and outputs signal components that are audibly unnecessary by the auditory masking effect among the signal components of the audio signal;
A speech encoding unit that performs speech encoding processing for compressing and encoding the output signal of the auditory masking shaping processing unit and outputting a bitstream;
A speech encoding apparatus comprising:   The speech coding according to claim 1, wherein the auditory masking shaping processing unit removes an audibly unnecessary frequency component on the frequency axis and outputs it back on the time axis for each predetermined time interval. apparatus. The auditory masking shaping processing unit
A frequency converter that converts the frequency of the block composed of the input audio signal;
A smoothing unit that smoothes the output signal of the frequency conversion unit;
A shaping unit that removes unnecessary frequency components in the output signal of the frequency conversion unit using the output signal of the smoothing unit as a masking threshold;
The speech coding apparatus according to claim 1, further comprising: a frequency inverse transform unit that inversely transforms an output signal of the shaping unit and outputs a shaped signal.   The speech coding apparatus according to claim 1, wherein the auditory masking shaping processing unit removes frequency components such that a predetermined number of frequency components on the frequency axis remain.   The speech coding apparatus according to claim 1, wherein the auditory masking shaping processing unit removes a frequency component in a predetermined band. Furthermore,
The speech coding apparatus according to any one of claims 1 to 5, further comprising a switching unit that analyzes characteristics of the input speech signal and switches whether to perform output via the auditory masking shaping processing unit.   The speech coding apparatus according to claim 6, wherein the switching unit selects an output to the auditory masking shaping processing unit when the input speech signal has a characteristic of a music signal.   The speech coding apparatus according to any one of claims 1 to 6, wherein the speech coding apparatus functions as a speech processing apparatus that distributes a music signal to a mobile phone terminal. Among the signal components of the input audio signal, the auditory masking effect suppresses signal components that are audibly unnecessary and outputs them.
Audio compression-coding the shaped signal in which the auditory unnecessary signal component is suppressed, and outputting a bitstream;
A speech encoding method characterized by the above.   The auditory unnecessary signal component is suppressed by removing an audibly unnecessary frequency component on the frequency axis and outputting it back on the time axis for each predetermined time interval. 10. The speech encoding method according to 9. The block composed of the input audio signal is frequency converted,
Smoothing the frequency converted signal;
Using the smoothed signal as a masking threshold, remove unnecessary frequency components from the frequency converted signal,
The speech encoding method according to claim 9 or 10, wherein the auditory unnecessary signal component is suppressed by inversely transforming the signal from which the unnecessary frequency component is removed.   The speech coding method according to any one of claims 9 to 11, wherein the audibly unnecessary signal component is suppressed by removing the frequency component so that a predetermined number of frequency components remain on the frequency axis.   The speech coding method according to any one of claims 9 to 12, wherein a signal component that is audibly unnecessary is suppressed by removing a frequency component of a predetermined band.   14. The auditory unnecessary signal component is suppressed after analyzing characteristics of an input audio signal and determining whether or not to suppress the auditory unnecessary signal component. The speech encoding method described in 1.   15. The speech encoding method according to claim 14, wherein when the input speech signal has characteristics of a music signal, the auditory unnecessary signal component is suppressed. A program to be executed by a computer constituting a speech encoding device,
Auditory masking shaping processing that suppresses and outputs signal components that are audibly unnecessary by the auditory masking effect among the signal components of the input audio signal;
A program that causes the computer to execute voice coding processing for voice compression coding the shaped signal subjected to the auditory masking shaping processing and outputting a bit stream.

Images