JP2004053763A - Speech encoding transmission system of multipoint controller - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize speech transmission of high quality even for simultaneous speaking of a plurality of speakers assumed in one-to-multiple or multiple-to-multiple teleconferencing and speaking in a highly-noisy environment. <P>SOLUTION: A speech encoding parameter control part 211 specifies one conference terminal decided as a voicing one among respective conference terminals to an exchange 201 and transmits some of a plurality of kinds of speech encoding parameters of a decoded speech signal of the conference terminal to the respective conference terminals connected to the exchange 201 without performing speech re-encoding processing. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a telephone conference (teleconference) using an encoded voice signal encoded based on an information source encoding scheme such as a code-excited linear prediction (hereinafter referred to as “CELP”) scheme. More specifically, the present invention relates to a multi-point control unit (hereinafter, referred to as “MCU”) applied to a system, and particularly to a multi-point control unit that re-encodes a decoded audio signal in response to a mixture of utterances by a large number of speakers. The present invention relates to a speech coded transmission system of a control device.
[0002]
[Prior art]
The MCU connects a plurality of communication terminals in various forms via a communication network, and receives and transmits information contents of different and different codes including video, audio, data, and the like handled by each communication terminal. It has been developed as an apparatus that provides a service of performing exchange and distribution processing in accordance with information content and delivering the processed information to a plurality of communication terminals. Typical applications are teleconferencing and videoconferencing systems. For audio transmission processing handled by the MCU, compression encoding is applied.
In recent years, as a technique for efficiently compressing and encoding voice in the telephone band, a technique based on an information source coding scheme such as the CELP scheme has been practically used mainly in fields such as digital mobile phones, international communications, and intra-company communications. Has been Among them, ITU-T Recommendation G. The CS-ACELP (Conjugate Structure-Algebraic CodeExcited Linear Prediction: Algebraic Code Excited Linear Prediction of Conjugate Structure) method used in G.729 and the GSM-AMR (Global System for Mobile-Regulation-Mime-Ratio-Mime-Ratio-Mobi-Ratio-Mobi-Ratio-Mobi. It has been adopted as a regional standard method.
[0003]
Among voice waves, especially voiced sounds are generated by adding resonance characteristics of a vocal tract such as an oral cavity and a nasal cavity to a vibration wave generated by vibration of a vocal cord. Originally, the CELP method is an encoding method that models such a human vocal mechanism. Here, the vocal cord vibration wave is represented by a pitch period representing a repetitive component thereof and a noise parameter representing a fluctuation component. In addition, the vocal tract transfer characteristics when the sound passes through the throat, mouth, and nose and the radiation characteristics of the lips are approximately expressed using a linear prediction method.
[0004]
In a specific CELP system, an adaptive codebook and a fixed codebook, which are basically two codebooks, are used, and an LSP quantized codebook (Line Spectral) is used. Pair quantification code book and gain quantization codebook are used. The adaptive codebook expresses a periodic signal component of a drive source signal, and a code that stores a previous drive source signal in a memory and is updated every frame period is often used. On the other hand, a random codebook expresses a non-periodic signal component that cannot be expressed by an adaptive codebook, and a fixed codebook of a plurality of typical signal patterns is often used. Since this type of noise codebook requires a large amount of memory to store a signal pattern, as an improved version, recently, an algebra that expresses an approximately aperiodic signal component with a small number of pulses has been developed. Codebooks have become popular. Further, the gain of the periodic / aperiodic component of the drive source signal is vector-quantized using a gain quantization codebook. Further, a linear prediction coefficient obtained as a result of the linear prediction analysis of the voice to be transmitted is converted into an LSP, and this is vector-quantized using an LSP quantization codebook.
[0005]
FIG. 16 is a block diagram showing a configuration example of a conventional speech coding apparatus based on the CS-ACELP system. In the figure, 604 is a linear prediction analysis processing unit, 605 is an LSP (Line Spectral Pair) quantization processing unit, 606 is an LSP quantization codebook, 608 is a multiplexing unit, 609 is an inverse quantization processing unit, and 610 is an adaptive code. Book, 611 is an algebraic codebook, 612 and 616 are adders, 613 is a gain control amplifier, 614 is a gain quantization codebook, 615 is a synthesis filter, 617 is an auditory weighting filter, 618 is a distortion minimizing section, and 624. Is an LPC (Linear Predictive Coding) LSP conversion unit.
[0006]
Next, the operation will be described.
The linear prediction analysis processing unit 604 obtains the LSP parameter from the speech input. The LSP parameter is directly input to the LSP quantization processing unit 605, and is encoded with reference to the LSP quantization codebook 606. The encoded LSP parameter (codebook index) is sent to the inverse quantization processing unit 609 and also to the multiplexing unit 608. The inverse quantization processing unit 609 calculates a linear prediction coefficient using the LSP coefficient obtained by referring to the LSP quantization codebook 606 based on the encoded LSP parameter (codebook index), and generates a synthesis filter 615. To supply. A processing block group including the synthesis filter 615 and including the adaptive codebook 610, the algebraic codebook 611, the adders 612 and 616, the gain control amplifier 613, the gain quantization codebook 614, and the perceptual weighting filter 617. By changing the combination of the adaptive codebook 610, the algebraic codebook 611, and the gain quantization codebook 614, a plurality of speech waveforms are synthesized. The distortion minimizing section 618 calculates an auditory weighted error power (= square error) between the plurality of synthesized speech waveforms and the input signal waveform, and among them, an adaptive codebook 610 and an algebraic codebook which minimize the error power. 611 and the combination of the gain quantization codebook 614 are selected. Speech coding processing is performed based on the so-called AbS (Analysis by Synthesis) method. The coding parameters (the adaptive codebook 610, the algebraic codebook 611, the gain quantization codebook 614, etc.) quantized in this way and the quantization parameters previously quantized by the LSP quantization processing unit 605 are: Are multiplexed by the multiplexing unit 608, and then sent to the decoder side.
[0007]
Next, a case will be described in which the above-described encoding method is applied to a telephone conference device using an MCU, and one-to-many or many-to-many communication is performed.
FIG. 17 is a block diagram showing the configuration of a conventional MCU. In FIG. 17, reference numeral 20 denotes a telephone constituting a conference terminal, reference numeral 21 denotes an exchange, reference numeral 22 denotes a line interface unit, reference numeral 23 denotes a voice decoding unit, and reference numeral 24 denotes a voice detection unit. , 25 is a noise suppressing unit, 26 is a voice adding unit, 27 is a distribution processing unit, 28 is a local terminal voice subtracting unit, 29 is an automatic gain control unit, and 30 is a voice re-encoding processing unit.
[0008]
Next, the operation will be described.
First, the coded voice received by the line interface unit 22 is decoded by the voice decoding processing unit 23 into a voice signal for each channel. For the decoded audio signal, the sound detection unit 24 detects the presence / absence of a sound / no sound for each from the conference terminal 20. The detection result is used for determining the addition target terminal in the voice addition unit 26, weighting for noise suppression in the noise suppression processing unit 25, and automatic adjustment of the voice level by the automatic gain control unit 29.
[0009]
When the number N of conference participants increases, the background noise level also increases in proportion to N, and therefore, there is a problem that a decrease in the S / N ratio causes deterioration in speech quality. Therefore, the noise suppression processing unit 25 sets each of the speech channel and the silence channel according to the speech / silence detection result of each conference participant in order to reduce the noise input from the channel in the silence state. Each weight coefficient is determined based on the number of conference participants N, and noise control processing is performed to multiply the decoded voice signal by the weight coefficient. The voice adding unit 26 adds the decoded voice signal of the channel to be added, which has been subjected to the noise control processing, with reference to the detection result of the voice detecting unit 24. The added decoded audio signal is redistributed to each channel in the distribution processing unit 27.
[0010]
The own-terminal audio subtraction unit 28 subtracts the own-terminal signal from the added decoded audio signal in order to suppress the echo caused by the rounding of the audio signal of the own terminal and eliminate annoying audibility. Further, in a multipoint telephone conference system, since there is a possibility that saturation distortion may be caused by adding voices of many people, the automatic gain control unit 29 prevents the saturation distortion, and furthermore, the individual and overall audio level To adjust. The output signal of the automatic gain control unit 29 is encoded by the audio re-encoding processing unit 30 and output to the line.
[0011]
A first problem of the system using the multipoint control device shown in FIG. 17 is that the voice quality is deteriorated due to repeated decoding and re-encoding (hereinafter, referred to as “tandem connection”). Since the audio encoding / decoding device based on the CELP method is lossy encoding, the decoded audio signal includes a quantization error. Furthermore, by connecting these in tandem, the quantization error is further accumulated in the re-encoding process, and the voice quality is degraded.
[0012]
As a method for solving this problem, for example, there is a technique disclosed in Japanese Patent Application Laid-Open No. 2000-174909 "Conference terminal control device". FIG. 18 is a block diagram showing a configuration of the multipoint control device (MCU). In the figure, 2a to 2m are demultiplexers, 4 is a speaker detection circuit, 6 is a first selector, 8a to 8n are decoders (speech decoding processing units), 10a to 10n are attenuation circuits, 12 is a synthesis circuit, and 14 is an encoder. (Audio re-encoding unit), 16 is a second selector, 18 is a distribution circuit.
[0013]
Next, the operation will be described.
When the MCU receives voice information from each conference terminal, voice / unvoiced signals are separated by the demultiplexers 2a to 2m, and the voice / unvoiced signal is used to determine which of the conference terminals from which the compressed voice code is voiced. The speaker detection circuit 4 detects it. In addition, the number of voiced conference terminals is measured. When a voiced or unvoiced determination is made, the information is sent to the first selector 6, and the first selector 6 operates as follows according to the number of conference terminals in a voiced state.
[0014]
If there are two or more voiced conference terminals, the first selector 6 selects the voiced conference terminals and connects them to the decoders 8a to 8n in a one-to-one relationship. Is sent. Each of the corresponding decoders 8a to 8n decodes the supplied compressed audio code to generate an audio signal. The generated audio signal is attenuated to a predetermined value by the attenuation circuits 10a to 10n. Each audio signal is subjected to level adjustment by the attenuation circuit 10 and then synthesized by the synthesis circuit. That is, all voices from the conference terminal in a voiced state are collected. The synthesized audio signal is re-encoded by the encoder 14, sent to the second selector 16, and transmitted by the distribution circuit 18 to all conference terminals.
[0015]
On the other hand, if there is only one conference terminal in the voiced state, the fact is transmitted to the first selector 6, the conference terminal in the voiced state is selected by the first selector 6, and the conference terminal is directly transmitted to the second selector 16 as it is. Connected. The compressed voice code of the conference terminal in a voiced state connected to the second selector 16 is transmitted to all conference terminals via the distribution circuit 18 as it is.
[0016]
When the MCU has the configuration shown in FIG. 18, in the case of a single speaker, all data of the coded voice is passed through, so that tandem connection is avoided and a low bit rate voice coding scheme can be used. High quality audio can be provided.
[0017]
As a second problem of the system using the multipoint control device shown in FIG. 17 described above, there is a deterioration factor peculiar to the information source coding process, that is, simultaneous utterances by a plurality of speakers. In this case, the deterioration is remarkable, and there is a problem that the hearing becomes hard to hear.
In the above-described conventional speech coding apparatus based on the CELP scheme, as shown in FIG. 16, coding is originally performed on the assumption that a single speaker speaks. That is, the vocal cord sound source (drive vector), vocal tract information (linear prediction coefficient), gain information, and the like are extracted in the encoding process and are quantized and transmitted. There is only one parameter indicating the feature amount of speech, and when transmitting mixed speech of a plurality of speakers, it cannot be encoded with high accuracy. For example, in the case of a mixed voice of a plurality of speakers, the vocal cord sound source includes a plurality of types of pitch period information different for each speaker, but there is no means for expressing the plurality of types of pitch periods. Also, in the case of a mixed voice of a plurality of speakers, the vocal tract information has a more complex spectral structure than that of a single speaker. Furthermore, a quantization pattern (quantization table) that faithfully expresses it is not prepared, and the error at the time of quantization tends to increase.
[0018]
As means for solving such a problem, for example, there is a method disclosed in Japanese Patent Application Laid-Open No. H10-240299, entitled "Speech Encoding and Decoding Device". FIG. 19 is a block diagram showing a configuration of a CELP speech encoding device disclosed in this publication. In the figure, 31 is a multi-speaker voice separation unit, 32 ₁ ~ 32 _N Is the long-term predictor, 33 ₁ ~ 33 _N Is a source codebook, 34 is a reflection coefficient analysis unit, 35 is a throat (throat) approximation filter, 36 ₁ ~ 36 _N , 37 are an adder, 38 is a subtractor, and 39 is an error analyzer.
[0019]
Next, the operation will be described.
The multi-speaker voice separating unit 31 analyzes the periodic characteristics of the input voice signal, specifies the number of speakers n (1 <n ≦ N), and separates the voice of each speaker included in the voice signal. Source voice A of each speaker ₁ ~ A _n Is output as The number n of speakers obtained by the multi-speaker voice separation unit 31 is supplied to the reflection coefficient analysis unit 34. In the reflection coefficient analysis unit 34, the reflection coefficient is an order corresponding to the number n of speakers, such as 10th order when the number n of speakers is one, 15th order when there are two speakers, and 20th order when it is more than two. Calculate r. The reflection coefficient r can be obtained, for example, by executing FLAT (fixed-point covariance lattice type algorithm) using autocorrelation of input speech. The obtained reflection coefficient r is given as a coefficient of the throat approximation filter 35.
[0020]
On the other hand, the source speech A of each speaker separated by the multi-speaker speech separation unit 31 ₁ ~ A _n Are n long-term predictors 32 ₁ ~ 32 _n Respectively. Long-term predictor 32 ₁ ~ 32 _n Then, these source sounds A ₁ ~ A _n Pitch L of the source voice from the correlation between the ₁ ~ L _n Is extracted. These pitches L ₁ ~ L _n And the source codebook 33 respectively ₁ ~ 33 _n From the adder 36 ₁ ~ 36 _n , And the source speech for each speaker is decoded. The source voices of the plurality of speakers are added by the adder 37, and the characteristics of the vocal tract are given by the throat approximation filter 35 to become a local decoded signal. The local decoded signal and the input voice are subtracted by a subtractor 38, and an error analyzer 39 generates a source codebook 33 so that an error signal from the subtracter 38 is minimized. ₁ ~ 33 _n Index I ₁ ~ I _n Are sequentially determined.
[0021]
[Problems to be solved by the invention]
Since the conventional voice coded transmission system of the multipoint control device is configured as described above, in order to avoid voice quality deterioration due to accumulation of quantization errors, when the method shown in FIG. 18 is used, There is a possibility that the speaker detection circuit 4 may react to short-period voices (such as coughing, hammering, etc.) uttered by a person other than the speaker, or cause erroneous determination due to a change in background noise. In this case, the circuit switching between pass-through and decoding / re-encoding processing is frequently performed. That is, in FIG. 18, the selector 6 switches between the output (encoded audio signal) of the encoder 14 and the encoded audio signal that has been passed through according to the determination result of the speaker detection circuit 4. However, in the case of a low-bit-rate audio coding method such as the CELP method, normally, a high-quality sound can be transmitted by always operating a pair of an encoder and a decoder. However, in the configuration of FIG. 18, by switching the selector 6, the encoder paired with the decoder built in the conference terminal of the receiver becomes the encoder built in the conference terminal of the sole speaker and the MCU. It is frequently switched to an encoder (encoder 14). For this reason, there is a problem on the terminal side that receives the sound that the sound becomes unnatural due to the discontinuity of the sound due to the switching and that the sound quality fluctuates frequently, which makes the sound hard to hear.
[0022]
Further, since the conventional voice coded transmission system of the multipoint control device is configured as described above, in order to avoid remarkable deterioration due to simultaneous utterances of a plurality of speakers, which is peculiar to the information source coding process, FIG. When the method described in (1) is used, the pitch period information must be prepared each time the number of speakers increases, and the bit rate increases in proportion to the increase in the number of speakers. There has been a problem that applicable transmission networks are limited, such as networks, for example, an asynchronous transmission network represented by an ATM network or an IP network.
[0023]
The present invention has been made in order to solve the above-mentioned problems, and has a simultaneous utterance of a plurality of speakers assumed in a one-to-many or many-to-many telephone conference, a utterance in a high noise environment, a coughing, It is an object of the present invention to obtain a voice coded transmission system of a multipoint control device capable of realizing high-quality voice transmission even for short sections such as a hammer.
It is another object of the present invention to provide a voice coded transmission system of a multipoint control device that can be applied to a transmission network in which the transmission speed is constant or the transmission speed per channel is limited.
[0024]
[Means for Solving the Problems]
A voice coded transmission system of a multipoint control device according to the present invention is connected to a plurality of communication terminals via a communication network, and receives and transmits a coded voice signal obtained by coding voice handled by each communication terminal. In the voice coded transmission system of the multi-point control device that performs predetermined processing according to the information content of these coded voice signals and distributes the processed coded voice signal to a plurality of communication terminals, Decoding the received encoded audio signal to generate a decoded audio signal, determining a communication terminal in which the decoded audio signal is voiced, specifying one of the communication terminals determined to be voiced, and specifying the specified communication terminal Some of the plurality of types of speech encoding parameters in the decoded speech signal of the communication terminal are transmitted to each communication terminal without performing speech re-encoding processing and designated. After performing a voice re-encoding process on the other voice coding parameters of the plurality of types of voice coding parameters in the decoded voice signal of the communication terminal and the plurality of types of voice coding parameters in the decoded voice signal of the other communication terminal. It is configured to transmit to each communication terminal.
[0025]
The voice coded transmission system of the multipoint control apparatus according to the present invention designates one communication terminal when only the decoded voice signal from one communication terminal has sound, and decodes the voice of the specified communication terminal. Some speech coding parameters in a signal are transmitted to each communication terminal without performing speech recoding processing.
[0026]
The voice encoding transmission system of the multipoint control device according to the present invention sets a priority order for a plurality of communication terminals, and when there are a plurality of communication terminals that have determined that the decoded voice signal is sound, the One of the communication terminals having the highest priority among the communication terminals is designated, and some of the speech encoding parameters in the decoded speech signal of the designated communication terminal are transmitted to each communication terminal without performing speech re-encoding processing. It is what was constituted.
[0027]
In the voice coded transmission system of the multipoint control device according to the present invention, when there are a plurality of communication terminals that have determined that the decoded voice signal has sound, the priority is set in the order of arrival of the voice signal first. It is configured as follows.
[0028]
The voice coded transmission system of the multipoint control device according to the present invention is configured so that one specific communication terminal is preferentially designated in advance and the priority is set.
[0029]
Some voice coding parameters in the voice coding transmission system of the multipoint control device according to the present invention are configured to be parameters carrying pitch period information.
[0030]
A part of speech encoding parameters in the speech encoding transmission system of the multipoint control device according to the present invention are configured to be parameters carrying spectrum envelope information.
[0031]
The voice coded transmission system of the multipoint control device according to the present invention, when it is determined that the decoded voice signals received from the plurality of communication terminals are voiced, the voice coding is performed according to the number of voiced communication terminals. It is configured to adaptively set the bit allocation of the frame of the parameterization parameter.
[0032]
In the voice coded transmission system of the multipoint control device according to the present invention, when the decoded voice signals received from the first and second communication terminals among the plurality of communication terminals are sound, If it is determined, the two speech coding parameters carry the pitch cycle information and the spectrum envelope information in the decoded speech signal of the first communication terminal and the pitch cycle information in the decoded speech signal of the second communication terminal. The voice coding parameter is transmitted to each communication terminal without performing the voice recoding process, and the voice coding parameter for performing the voice recoding process on the decoded voice signals of the first and second communication terminals is It is configured to perform degenerate quantization bit control based on a transmission rate control function, and then transmit to each communication terminal.
[0033]
The voice coded transmission system of the multipoint control apparatus according to the present invention is configured such that, when it is determined that the decoded voice signal received from the main voice communication terminal and the sub voice communication terminal among the plurality of communication terminals is sound, The two speech coding parameters that carry the pitch period information and the spectrum envelope information in the decoded speech signal of the main speech communication terminal and the speech coding parameter that carries the pitch period information in the decoded speech signal of the sub speech communication terminal Speech encoding parameters to be transmitted to each communication terminal without performing re-encoding processing and subjected to audio re-encoding processing in the decoded audio signals of the main audio and sub audio communication terminals are degenerated quantum based on the transmission rate control function. This is configured to transmit to each communication terminal after performing coded bit control.
[0034]
The encoding parameter for performing the degenerate quantization bit control in the speech encoding transmission system of the multipoint control device according to the present invention is configured to be a gain codebook.
[0035]
The encoding parameter for performing the degenerate quantization bit control in the speech encoding transmission system of the multipoint control apparatus according to the present invention is configured to be a noise codebook.
[0036]
The voice coded transmission system of the multipoint control device according to the present invention may be configured such that voice coded parameters are subjected to voice re-coding processing and then transmitted to each communication terminal or each communication terminal is not subjected to voice re-coding processing. The encoding mode information for deciding whether to transmit the data is transmitted in a frame of an encoding parameter having a predetermined number of bits.
[0037]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of a multipoint control apparatus to which a voice coded transmission system according to Embodiment 1 of the present invention is applied. In FIG. 1, 201 is an exchange, 202 is a line interface unit, 203 is a speech decoding unit, 204 is a speech detection unit, 205 is a noise suppression processing unit, 206 is a speech addition unit, 207 is a distribution processing unit, and 208 is a local terminal. An audio subtraction unit, 209 is an automatic gain control unit, 210 is an audio re-encoding processing unit, and 211 is an audio encoding parameter control unit.
[0038]
The basic functions of the configuration of FIG. 1 are as follows.
Coded voice from each conference terminal (communication terminal) is received by the line interface unit 202 via the exchange 201. The received encoded speech is decoded by the speech decoding processing unit 203 into a speech signal for each channel. The sound detection unit 204 detects the presence / absence of a sound / no sound in each of the decoded audio signals from the conference terminal. This detection result is used for determining the addition target terminal in the voice addition unit 206, weighting for noise suppression in the noise suppression processing unit 205, and automatic adjustment of the voice level by the automatic gain control unit 209. Further, the decoded audio signal decoded by the audio decoding processing unit 203 and the extracted audio parameters are input to the audio encoding parameter control unit 211.
[0039]
When the number N of conference participants increases, noise increases and the communication quality deteriorates. Therefore, the noise suppression processing unit 205 reduces noise input from a channel in a silent state. For this purpose, each weighting factor based on the number N of conference participants is determined for each of the sound channel and the silent channel in accordance with the result of detection of voice / non-voice of each conference participant, and this weighting factor is determined for each decoded speech signal. To perform noise control processing. The voice adding unit 206 adds the decoded voice signal of the channel to be added, which has been subjected to the noise control processing, with reference to the detection result of the voice detecting unit 204. The added decoded audio signal is redistributed to each channel in distribution processing section 207.
[0040]
The own-terminal audio subtraction unit 208 subtracts the own-terminal signal from the added decoded audio signal in order to suppress the echo caused by the rounding of the audio signal of the own terminal and eliminate the annoying audibility. Further, in a multipoint telephone conference system, since there is a possibility of causing a saturation distortion due to the addition of voices of a large number of people, the automatic gain control unit 209 prevents the saturation distortion, and furthermore, the individual and overall voice level. To adjust. The output signal of the automatic gain control unit 209 is encoded by the audio re-encoding processing unit 210 and output to the line.
[0041]
FIG. 2 is a block diagram showing a more detailed configuration of the speech coding parameter control unit 211 of FIG. 1. In the figure, 212 is a selector, 213 is a distribution processing unit, and 214 is a speaker selection unit.
The basic functions of the configuration of FIG. 2 are as follows.
The speaker selection unit 214 adds up the detection results of the voice detection unit 204 of each channel, and when it can be considered that the speaker is the only one, the information (such as a channel number) of the speaker is selected by the selector 212 and the voice recoding. Output to the conversion processing unit 210. The selector 212 selects an encoding parameter to be passed through to the audio re-encoding processing unit 210 from the audio encoding parameters of each channel extracted by the audio decoding processing unit 203, according to the selection result of the speaker selection unit 214. . The distribution processing unit 213 redistributes the encoding parameter selected by the selector 212 to the audio re-encoding processing unit 210 of each channel.
[0042]
FIG. 3 is a block diagram showing a more detailed configuration of the speech decoding processing unit 203 of FIG. 1. In the drawing, 126 is a demultiplexing unit, 131a is an adaptive codebook, 131b is a gain decoding unit, and 132 is a decoding gain MA. A prediction unit, 133 is an algebraic code decoding unit, 134 is a pitch prefilter, 135 is an LSP decoding unit, 136 is an LSP interpolation unit, 137 is an LSP / LPC conversion unit, 127 and 128 are control amplification units, 129 is an adder, 138 is a synthesis filter, and 139 is a post filter.
[0043]
FIG. 4 is a block diagram illustrating a configuration of the audio re-encoding processing unit 210 in FIG. 1, and corresponds to the above-described audio encoding device in FIG. In the figure, reference numeral 104 denotes a linear prediction analysis processing unit, 105 denotes an LSP (Line Spectral Pair) quantization processing unit, 106 denotes an LSP quantization codebook, 108 denotes a multiplexing unit, 109 denotes an inverse quantization processing unit, and 110 denotes an adaptive code. Book, 111 is an algebraic codebook, 112 and 116 are adders, 114 is a gain quantization codebook, 115 is a synthesis filter, 117 is an auditory weighting filter, 118 is a distortion minimizing unit, and 119, 120 and 122 are changeover switches. , 113a and 113b are gain control amplification units, 121 is a pitch pre-filter, 125 is a gain MA prediction unit, and 124 is an LPC (Linear Predictive Coding) / LSP conversion unit.
[0044]
Next, the operation will be described. Note that the speech coding scheme used for the description is described in ITU-T Recommendation G. 729 CS-ACELP.
The audio decoding processing unit 203 executes a decoding process based on the transmitted audio encoded data. At the same time, a parameter indicating a characteristic amount unique to the speaker's voice, that is, an adaptive codebook index which is pitch period information expressing a repetition period of a vocal cord vibration wave, and an LSP which is spectrum envelope information expressing vocal tract information. The codebook index is output to the speech coding parameter control unit 211. In this case, if the speaker is uniquely determined by the determination result of the voice detection unit 204, the adaptive codebook index and the LSP codebook among the encoding parameters received from the channel allocated to the terminal device of the speaker are used. The index and the index are directly passed through to the audio re-encoding processing unit 210.
[0045]
At this time, the voice re-encoding processing unit 210 connects the changeover switches 119, 120, and 122 to the contacts 119A, 120A, and 122A, and performs the re-encoding process, that is, the code book search process for the passed-through encoding parameters. Instead, it is sent to multiplexing section 108 as it is. For the other parameters (algebraic codebook index and gain codebook index in FIG. 4), re-encoding processing is performed based on the decoded speech signal to which speech has been added, and distortion minimizing section 118 searches for a least square error. To extract the optimum quantization value and send it to the multiplexing unit 108. The multiplexing unit 108 multiplexes these parameters and outputs them to the line interface unit 202.
[0046]
Here, regarding the voice detection unit 204, for example, when a voice is interrupted from another channel while a voice is being voiced on a certain channel, even if a rising edge of voice is detected on that channel, the voice detection unit 204 immediately receives a voice from the currently voiced channel. The switching is not performed, and a predetermined waiting time is given to delay the switching from the channel in which the speech is being made, thereby preventing the switching in the less important speech of a relatively short section, such as hammering and coughing. If this is a relatively long interrupt message, a switching delay occurs, but this is reflected in the output (addition decoded voice signal) of the voice addition unit 206, so that the voice re-encoding processing unit 210 Since it is encoded, there is no worry about running out of head. However, since the adaptive codebook index (pitch period information) and the LSP codebook index (spectral envelope information), which are encoding parameters including important information for decoding a speech signal, are partially missing, the original speaker However, the quality of the interrupt message has deteriorated. However, in the operation of the conference, the importance of interrupt remarks is often low, and information about frequency components is also leaked in the algebraic codebook index, albeit slightly. However, the sound of the interrupting speaker, such as decoding of unpleasant sounds, does not become abnormal, and it can be said that this deterioration is relatively unnoticeable.
[0047]
In the case where the speaker is not uniquely determined in the voice detection unit 204, the pass-through operation is performed by connecting the switches 119, 120, and 122 of the voice re-encoding processing unit 210 to the terminals 119B, 120B, and 122B, respectively. Not performed. Therefore, predetermined re-encoding processing is performed in the audio re-encoding processing section 210 for the decoded audio of each channel having a speaker input.
[0048]
As described above, according to the first embodiment, some of the plurality of types of speech encoding parameters in the decoded speech signal of the designated conference terminal are not subjected to speech re-encoding processing. Among the plurality of types of speech encoding parameters transmitted to each communication terminal and in the decoded speech signal of the designated conference terminal, other speech encoding parameters, and a plurality of kinds of speech encoding parameters in the decoded speech signal of another communication terminal Is transmitted to each communication terminal after performing voice re-encoding processing, so simultaneous speeches of multiple speakers assumed in a one-to-many or many-to-many conference call, speeches in a high noise environment, It is possible to obtain an effect that high-quality voice transmission can be realized even for short sections such as coughing and companion speech.
[0049]
That is, by passing through some of the voice coding parameters, the main speaker (the speaker of the designated conference terminal) deteriorates by repeating decoding and re-encoding for the main voice coding parameters. Therefore, the main speaker can be reproduced with a sound quality close to the sound quality at the time of pass-through. Further, since the encoding parameter always passes through the audio re-encoding processing unit, even if frequent switching occurs, the effect of eliminating discontinuity in audio can be obtained. In addition, since the quality is always kept constant, an effect is obtained that the unnatural feeling due to the fluctuation of the voice quality is eliminated.
[0050]
Further, according to the first embodiment, some of the speech encoding parameters that pass through without performing speech re-encoding processing are parameters that carry pitch period information, and therefore, vocal cord vibrations specific to the speaker's speech. An effect is obtained in that information expressing a number of repetition periods is transmitted as it is, and high-quality voice transmission can be realized.
[0051]
Also, according to the first embodiment, some of the speech encoding parameters that pass through without performing speech re-encoding processing are parameters that carry spectral envelope information, and therefore, the voices of the speaker's oral cavity, nasal cavity, etc. An effect is obtained that the information expressing the road can be transmitted as it is to realize high quality voice transmission.
[0052]
Embodiment 2 FIG.
In the second embodiment, the configuration corresponding to FIGS. 1 to 3 of the first embodiment is almost the same. However, in the second embodiment, the adaptive codebook index that is passed through from speech decoding processing section 203 to speech encoding parameter control section 211 contains the encoding parameters of the speaker of the conference terminal having the first priority. The first-order adaptive codebook index and the second-order adaptive codebook index among the encoding parameters from the speaker of the conference terminal having the second priority. FIG. 5 is a block diagram showing a configuration of a sound reproduction / coding processing unit according to Embodiment 2 of the present invention. In the drawing, 130 is a changeover switch, 110b is a second adaptive codebook, and 113c is a gain control amplification unit. , 140 are coding rate control units. The other components are the same as the components of the speech re-encoding processing unit according to Embodiment 1 shown in FIG. 4, and thus are denoted by the same reference numerals, and description thereof will be omitted in principle.
[0053]
FIG. 6 is a block diagram showing the configuration of a telephone or conference terminal connected to the exchange 201. In the figure, reference numeral 500 denotes a line interface unit, 501 denotes a voice decoding processing unit, 502 denotes a D / A converter, and 503 denotes a speaker. is there.
[0054]
FIG. 7 is a block diagram showing a more detailed configuration of the audio decoding processing unit 501 in the conference terminal shown in FIG. 6. In the drawing, 504 is a demultiplexing unit, and 505 is an encoding mode decoding unit. Other components are the same as some of the components included in FIG. 5, and thus are denoted by the same reference numerals, and description thereof will be omitted in principle.
[0055]
FIG. 8 is an explanatory diagram showing an example of a frame configuration of encoding parameters in the case of one speaker, the case of two speakers, and the case of three speakers.
[0056]
Next, the operation will be described. Note that, as in the case of the first embodiment, the speech coding method used for the description is described in ITU-T Recommendation G. 729 CS-ACELP.
Switch 130 switches between inputting and not inputting the second-order adaptive codebook index from speech coding parameter control section 211. The encoding rate control unit 140 is provided between the audio encoding parameter control unit 211 and the audio re-encoding processing unit 210, and encodes parameters based on the speaker selection information from the audio encoding parameter control unit 211. A control signal for determining the rate is output to algebraic codebook 111 and gain quantization codebook 114, and coding mode information is given as a switching control signal for switches 119, 120, and 130, and is output to multiplexing section 108. The demultiplexing unit 504 separates the speech coding parameters multiplexed by the multiplexing unit 108 in FIG. The encoding mode decoding unit 505 decodes the encoding mode information (the value of the last one bit in the frame in FIG. 8) separated by the demultiplexing unit 504.
[0057]
Hereinafter, the overall operation will be described with reference to FIGS.
Now, when there is only one speaker, the speaker selection unit 214 of FIG. 2 determines “one-speaker” based on the detection result of the voice detection unit 204 of FIG. Is output to the selector 212 and transferred to the voice re-encoding processor 210. The selector 212 selects an encoding parameter received from a channel corresponding to the conference terminal of one speaker, and outputs the selected encoding parameter to the distribution processing unit 213. The distribution processing unit 213 passes the encoding parameters (the LSP encoding index and the first-order adaptive encoding index) to the audio re-encoding processing unit 210 as it is. The operation at this time is to connect the switch 130 to the 130B side. That is, the input of the second adaptive codebook index is turned off. Further, the bit rate output from coding rate control section 140 is determined. Other operations are exactly the same as those in the first embodiment.
[0058]
Also, when there is no speaker or three or more speakers speak at the same time, the speaker selection unit 214 determines that “three or more speakers” is determined based on the detection result of the voice detection unit 204. Then, the speaker selection information indicating the determined content is output to the selector 212 and transferred to the voice re-encoding processing unit 210. In this case, the selector 212 does not select an encoding parameter. The voice re-encoding processing unit 210 performs a re-encoding process by tandem connection according to the speaker selection information. That is, the switches 119, 120, and 130 are connected to the contacts 119B, 120B, and 130B, respectively, so that the input of the first and second adaptive codebook indexes is turned off without passing through the LSP coding index. . Then, re-encoding processing is performed on all the encoding parameters based on the signal subjected to the voice addition, and an optimum quantization value is searched for and sent to the multiplexing unit 108. However, if there is no speaker, no re-encoding process is performed because there is no encoding parameter.
[0059]
On the other hand, when two speakers speak at the same time, the speaker selection unit 214 determines that the message is “two speakers”, and outputs speaker selection information indicating the determined content to the selector 212. The data is transferred to the audio re-encoding unit 210. The selector 212 selects the encoding parameter received from the channel corresponding to the conference terminal of the two speakers and outputs the selected encoding parameter to the distribution processing unit 213. The distribution processing unit 213 passes the encoding parameter to the audio re-encoding processing unit 210 as it is. The speech re-encoding processing unit 210 passes through the LSP encoding index of the first speaker and the first adaptive encoding index of the first speaker, and also passes through the second adaptive encoding index of the second speaker (pitch period). Information).
[0060]
That is, the switches 119, 120, and 130 are connected to the contacts 119A, 120A, and 130A, respectively, and the LSP coding index and the first and second adaptive codebook indexes are sent to the multiplexer 108. The other coding parameters, ie, the algebraic coding index and the gain coding index, are re-encoded based on the signal to which the voice is added, search for the optimal quantization value, and send it to the multiplexing unit 108. The multiplexing unit 108 multiplexes the plurality of coding parameters and outputs the multiplexed coding parameter to the line interface unit 202.
[0061]
At the same time, the coding rate control unit 140 adjusts the number of bits allocated to the algebraic coding index and the gain coding index according to the speaker selection information from the voice coding parameter control unit 211 so as to match the transmission rate. Then, while outputting to the algebraic codebook 111 and the gain quantization codebook 114, respectively, it generates coding mode information to be described later and outputs it to the switches 119, 120, and 130 and the multiplexing unit 108.
[0062]
In this case, the encoding mode is set to two modes, the encoding mode of "one speaker's speech" and "three or more speakers 'is set to" mode 0 ", and the encoding mode of" two speakers' speech " Is "mode 1". That is, the coding mode information of “0” and “1” is set. Therefore, one bit is required to transmit the encoding mode information. FIGS. 8 and 9 show an example of bit allocation of one frame (80 bits) of the encoding parameter in each mode when the transmission speed is 8 kbit / sec.
[0063]
In “mode 0”, ITU-T Recommendation G. This is almost the same as the bit assignment shown in FIG. That is, as shown in FIG. 8, in the case of “one-speaker utterance”, the coding parameters of the LSP codebook index and the adaptive codebook index are passed through, and in the case of “three-or-speaker utterance”, all Do not pass through the encoding parameters of However, since the transmission speed is 8 kilobits / second, there is no room for transmitting the encoding mode information in the standard system. Therefore, as shown in FIG. 9, one bit set as a parity bit in the standard system is encoded. Divert to mode information and transmit.
[0064]
On the other hand, in “mode 1”, the second-order adaptive coding index (13 bits) needs to be transmitted, so the bit allocation of other coding parameters is reduced by 13 bits (this is referred to as degeneration). There is a need. Therefore, as shown in FIGS. 9 and 8, in the case of “two-speaker”, the algebraic codebook index is reduced from 34 bits to 22 bits, and the gain codebook index is reduced from 14 bits to 13 bits. . Then, it passes through an encoding parameter consisting of the LSP codebook index and the adaptive codebook index of the first speaker, and the second adaptive codebook index of the second speaker.
[0065]
In addition, regarding the degeneracy method of the algebraic codebook index, the ITU-T Recommendation G. 729 Annex D This is realized by using the method (11-bit quantization) described in section 5.8. As for the degeneration method of the gain codebook index, the first subframe is defined in ITU-T Recommendation G. Using the method (7-bit quantization) described in section 3.9 of the G.729 body, the second subframe is defined in ITU-T Recommendation G.323. 729 Annex D This is realized by using the method (6-bit quantization) described in section 5.9.
[0066]
The multiplexing unit 108 multiplexes these coding parameters, and transmits the multiplexed parameters to each conference terminal via the line interface unit 202 and the exchange 201 for each frame shown in FIG. Then, in the receiving unit of the conference terminal shown in FIG. 6, speech decoding processing unit 501 decodes the frame of the encoding parameter received via exchange 201 and line interface unit 500. That is, in speech decoding processing section 501 shown in FIG. 7, demultiplexing section 504 separates one frame of encoding parameters into respective encoding parameters, and outputs encoding mode information to encoding mode decoding section 505. The coding mode decoding unit 505 provides a switch control signal to the switch 130 based on the coding mode information, and provides bit allocation information to the algebraic codebook 111 and the gain quantization codebook 114.
[0067]
Therefore, when a frame whose encoding mode information is 0 is received, the switch 130 is connected to the contact 130B side by the switching control signal, and turns off the input of the second adaptive codebook index. In addition, the algebraic codebook 111 uses ITU-T Recommendation G. The algebraic codebook index is decoded using the decoding method described in section 4.1 of the 729 body. In addition, the gain quantization codebook 114 also complies with ITU-T Recommendation G. The gain codebook index is decoded using the decoding method described in section 4.1 of the G.729 body.
[0068]
On the other hand, when the frame of which the encoding mode information is 1 is received, the switch 130 is connected to the contact point 130A side by the switching control signal, and turns on the input of the second adaptive codebook index to start decoding. In addition, the algebraic codebook 111 determines that the first subframe according to ITU-T Recommendation G. The algebraic codebook index is decoded using the decoding method described in section 4.1 of the G.729 main body, and the second subframe is described in Recommendation G.729. 729 Annex D. The algebraic codebook index is decoded using the decoding method described in Chapter 6. Next, the pitch period component of the second speaker obtained by decoding the second adaptive codebook index and the pitch period component of the first speaker obtained by decoding the first adaptive codebook index , And the noise component obtained by decoding the algebraic codebook index are added by the adder 112 and output to the synthesis filter 115 as an excitation signal. The synthesis filter 115 convolves the vocal tract information based on the excitation signal to obtain a decoded voice.
[0069]
As described above, according to the second embodiment, when it is determined that the decoded audio signals received from the plurality of conference terminals are voiced, the encoding parameter is determined according to the number of voiced conference terminals. Since the bit allocation of the frame is adaptively set, an effect is obtained that the present invention can be applied to a transmission network where the transmission speed is constant or the transmission speed per channel is limited.
[0070]
Further, according to the second embodiment, the decoded voice signals received from the first speaker of the conference terminal having the first priority and the second speaker of the conference terminal having the second priority among the plurality of conference terminals are provided. If it is determined that there is speech, a speech coding parameter consisting of a first-order adaptive codebook index, which is the pitch period information in the decoded speech signal of the first speaker, and an LSP codebook index, which is the spectrum envelope information. , And a speech encoding parameter consisting of a second-order adaptive codebook index, which is pitch period information in the decoded speech signal of the second speaker, is passed through to each conference terminal without performing speech re-encoding processing. Since the transmission is performed, even if two speakers speak at the same time, an effect of enabling transmission of relatively good voice quality can be obtained.
[0071]
Further, according to the second embodiment, since the degenerate quantization bit control based on the transmission rate control function is performed on the voice coding parameter for performing the voice re-encoding process, the voice quality is not affected by the deterioration of the voice quality. Each of the conference terminals is degenerated by reducing a small number of encoding parameters, and performing a speech re-encoding process on a speech encoding parameter including a second-order adaptive codebook index which is pitch period information in a decoded speech signal of a second speaker. , The effect of suppressing deterioration of voice quality can be obtained even when two speakers speak at the same time.
[0072]
Further, according to the second embodiment, since the coding parameter for which the degenerate quantization bit control is performed is the gain codebook index which is the excitation gain, the gain codebook index which has little effect on the deterioration of voice quality Thus, even if two speakers speak at the same time, it is possible to suppress the deterioration of voice quality.
[0073]
Further, according to the second embodiment, since the coding parameter for performing the degenerate quantization bit control is the algebraic codebook index which is a noise codebook, the algebraic codebook having little effect on the deterioration of speech quality. An effect is obtained in which even if two speakers speak at the same time by degenerating the encoding parameter of the index, deterioration in voice quality can be suppressed.
[0074]
Further, according to the second embodiment, the audio coding parameter is transmitted to each conference terminal after performing the audio re-encoding process, or is passed through each conference terminal without performing the audio re-encoding process. Since 1-bit coding mode information for determining whether to transmit is included in a frame of 80-bit coding parameters, information on whether to pass through a specific coding parameter without affecting the transmission rate. Can be included in the frame of the encoding parameter.
[0075]
Embodiment 3 FIG.
FIG. 10 is a block diagram illustrating a configuration of a multipoint control apparatus according to Embodiment 3 of the present invention, and FIG. 11 is a block diagram of a speech coding parameter control unit according to Embodiment 3. 10 corresponding to FIG. 1 are denoted by the same reference numerals, and portions of FIG. 11 corresponding to FIG. 2 are denoted by the same reference numerals, and the description thereof is omitted in principle. In FIGS. 10 and 11, reference numeral 215 denotes a first-arrival channel determination unit which responds to a detection result of a channel first detected by the voice detection unit 204 among channels in which speech has been made, and controls the selector 212 of the voice-coding parameter control unit 211 in response to the detection result. Control.
[0076]
Next, the operation will be described.
When the speakers compete, the first-arrival channel determination unit 215 determines the channel that has uttered first as the priority speaker, and provides the determination result to the selector 212. The selector 212 passes through only the encoding parameter of the first channel to the audio re-encoding processing unit 210. After that, the audio re-encoding processor 210 operates in the same manner as in the first embodiment.
[0077]
As described above, according to the third embodiment, priorities are set for a plurality of conference terminals, and when there are a plurality of conference terminals that have determined that the decoded audio signal is a sound, among them, Since one conference terminal having the highest priority is designated, and some of the speech encoding parameters in the decoded speech signal of the designated conference terminal are transmitted to each communication terminal without performing speech re-encoding processing, a plurality of This makes it possible to avoid confusion in the conference even when the speakers of the conference terminal speak simultaneously.
[0078]
Further, according to the third embodiment, if there are a plurality of conference terminals that have determined that the decoded audio signal is sound, priority is set in the order of arrival of the audio signal first, so that unnecessary This has the effect that the progress of the conference is not disturbed by coughing, a hammering, or an interrupted speech that deviates from the rules.
[0079]
Embodiment 4 FIG.
FIG. 12 is a block diagram showing a multipoint control device and another configuration according to Embodiment 4 of the present invention, and FIG. 13 is a block diagram of a speech coding parameter control unit according to Embodiment 4. 12 corresponding to FIG. 1 are denoted by the same reference numerals, and those in FIG. 13 corresponding to FIG. 2 are denoted by the same reference numerals, and description thereof is omitted in principle. In FIG. 12, reference numeral 224 denotes an MCU control unit which registers a specific channel as a priority speaker through the Internet or the like in the priority speaker determination unit 216. In FIG. 13, reference numeral 216 denotes a priority speaker determination unit, in which a specific channel of the telephone conference is registered in advance as a priority speaker, and a channel determined to be sound based on the decoded voice signal and the voice detection result is registered. If the channel is a specific channel, the speaker of that channel is determined as the priority speaker.
[0080]
Next, the operation will be described.
For example, the conference organizer registers the channel of the conference organizer or the channel of the designated moderator via the MCU control unit 224 via the Internet or the like at the time of setting the conference as the priority speaker and registers it in the MCU control unit 224. Keep it. Next, in a case where the speakers compete with each other, when there is a speaker on the channel of the registered priority speaker, that is, the channel of the conference organizer or the channel of the designated moderator, the priority speaker determination unit 216 performs decoding. The channel of the priority speaker is detected from the voice signal, the voice detection result, and the registered data, and the selector 212 of the voice coding parameter control unit 211 determines only the coding parameter of the corresponding channel by the voice recoding processing unit 210. Control to pass through to. Subsequent operations are the same as in the first embodiment.
[0081]
As described above, according to the fourth embodiment, one specific conference terminal is specified preferentially in advance, and the specific conference terminal is included in the plurality of conference terminals that have been determined to have sound in the decoded audio signal. Is included, a part of the speech encoding parameters in the decoded speech signal of the specific conference terminal is transmitted through each conference terminal without performing speech re-encoding processing, so that the conference host In this case, the speech of the speaker or the nominating facilitator is detected, and even if the speakers compete with each other, the effect is obtained that the conference can be smoothly performed.
[0082]
Embodiment 5 FIG.
FIGS. 14 and 15 are explanatory diagrams showing the frame configuration of the encoding parameters according to Embodiment 5 of the present invention. Note that, in the fifth embodiment, the configurations of the multipoint control device, the audio re-encoding processing unit in the multipoint control device, and the audio decoding processing unit in the conference terminal are shown in FIGS. This is the same as the configuration in the second embodiment shown in FIG. The configuration of the speech coding parameter control unit in the multipoint control device is the same as the configuration in Embodiment 4 shown in FIG.
[0083]
Next, the operation will be described. As in the second embodiment, the speech coding method used for the description is also the same as in the second embodiment. 729 CS-ACELP system. In FIG. 1, according to the detection result of the voice detection unit 204, the selection of the speaker selection unit 214 of the voice coding parameter control unit 211 in FIG. If it is determined to be “speaker or more”, as in the second embodiment, the speaker selection information based on the determination is provided to the selector 212 and output to the coding rate control unit 140 in FIG. The encoding rate control unit 140 generates 1-bit encoding mode information indicating “mode 0” or “mode 1” according to the speaker selection information, and outputs the generated encoding mode information to the audio re-encoding processing unit 210. . The speech re-encoding processing unit 210 forms a frame of the encoding parameter to be output from the multiplexing unit 108 to the conference terminal based on the encoding mode information.
[0084]
That is, in “mode 1”, since the second adaptive coding index (8 bits) needs to be transmitted, it is necessary to degenerate other coding parameters by 8 bits. Therefore, as shown in FIGS. 14 and 15, in the case of “two-speaker's speech”, the algebraic codebook index is reduced from 34 bits to 28 bits, and the gain codebook index is reduced from 14 bits to 12 bits. . Then, it passes through an encoding parameter consisting of the LSP codebook index and the adaptive codebook index of the first speaker, and the second adaptive codebook index of the second speaker.
[0085]
As for the degeneration method of the algebraic codebook index, the ITU-T Recommendation G. 729 Annex D The method (11-bit quantization) described in section 5.8 is used, and the ITU-T recommendation G. This is realized using the method (17-bit quantization) described in section 3.8 of the G.729 body. As for the degeneration method of the gain codebook index, both the first subframe and the second subframe conform to ITU-T Recommendation G.264. 729 Annex D This is realized by using the method (6-bit quantization) described in section 5.9.
[0086]
By the way, channels used in the telephone conference system are composed of a main audio and a sub audio. In this case, the main voice is the voice of the speaker, and the sub-voice is, for example, a simultaneous translation for the voice of the speaker. Therefore, in the priority speaker determination unit 216 in FIG. 13, the ranking is set in such a manner that the main voice is the first speaker and the auxiliary voice is the second speaker. As shown in FIG. 14, for the LSP codebook index and the first-order adaptive codebook index which are the coding parameters of the main voice, the quantized parameters passed through are transmitted as they are, For the second-order adaptive codebook index of, the quantization parameter that has been passed through is degenerated and transmitted.
[0087]
As described above, according to the fifth embodiment, when it is determined that the decoded audio signals received from the main audio and sub audio conference terminals are sound, the decoded audio signals of the main audio conference terminal are determined. Two speech coding parameters each carrying pitch cycle information and spectrum envelope information, and speech coding parameters carrying pitch cycle information in a decoded speech signal of a sub-speech conference terminal, are not subjected to speech re-encoding processing. Speech encoding parameters to be transmitted to the communication terminal and subjected to audio re-encoding are transmitted to each conference terminal after performing degenerate quantization bit control based on the transmission rate control function, so that the main audio and the sub audio are Even when speaking at the same time, the coding parameters of the sub-speech are degenerated, minimizing the deterioration of the speech quality of the main speech while maintaining the transmission rate. Effect of transmission is obtained.
[0088]
Further, according to the fifth embodiment, similarly to the second embodiment, when the coding parameters of the gain codebook index which have little effect on the deterioration of voice quality are degenerated, and two speakers speak simultaneously However, the effect that the deterioration of the voice quality can be suppressed can be obtained. In addition, it is possible to obtain an effect that the coding parameters of the algebraic codebook index having little effect on the deterioration of the voice quality are degenerated, and the deterioration of the voice quality can be suppressed even when two speakers speak simultaneously. Further, an effect is obtained that the encoding mode information indicating whether or not to pass through a specific encoding parameter can be included in the frame of the encoding parameter without affecting the transmission rate.
[0089]
Further, according to the second to fifth embodiments, as in the first embodiment, simultaneous utterances of a plurality of speakers assumed in a one-to-many or many-to-many telephone conference, or in a high-noise environment. This makes it possible to achieve high-quality voice transmission even for short-term speeches such as speeches, coughing, and hammering. Further, there is an effect that information expressing the repetition period of the vocal cord frequency unique to the speaker's voice is transmitted as it is, and high-quality voice transmission can be realized. Further, there is an effect that information expressing the vocal tract such as the oral cavity and nasal cavity of the speaker is transmitted as it is, and high-quality voice transmission can be realized.
[0090]
Note that, in each of the above embodiments, the present invention has been described using a conference terminal (telephone) as an example of a communication terminal, but the form of the communication terminal is not limited to the conference terminal. For example, in a broadcasting system or a cable broadcasting system that simultaneously transmits a plurality of types of audio signals on different channels, a plurality of receivers that receive the audio signals are applied as communication terminals, and a specific one channel (for example, , A main audio channel or a sub audio channel).
[0091]
【The invention's effect】
As described above, according to the present invention, the voice coded transmission system of the multipoint control device decodes the coded voice signal received from each communication terminal to generate a decoded voice signal, and the decoded voice signal is Is determined, and one of the communication terminals determined to have sound is designated, and some of the speech encoding parameters of a plurality of types of speech encoding parameters in the decoded speech signal of the designated communication terminal are determined. Is transmitted to each communication terminal without performing voice re-encoding processing, and among a plurality of types of voice coding parameters in the decoded voice signal of the designated communication terminal, other voice coding parameters and decoded voice of another communication terminal Since a plurality of types of voice coding parameters in a signal are transmitted to each communication terminal after performing voice recoding processing, a one-to-many or many-to-many telephone conference A plurality of speakers simultaneously speaking of contemplated and, speaking in a high noise environment, throat clearing, there is an effect that it also provides high-quality voice transmission against remarks short interval, such as back-channel feedback.
[0092]
According to the present invention, the speech coded transmission system of the multipoint control device designates one communication terminal when only the decoded speech signal from one communication terminal has sound, and specifies the communication terminal of the designated communication terminal. Since a part of the voice coding parameter in the decoded voice signal is transmitted to each communication terminal without performing voice re-encoding processing, high quality voice transmission in a one-to-many or many-to-many telephone conference is performed. There is an effect that can be realized.
[0093]
According to the present invention, the voice coded transmission system of the multipoint control apparatus sets a priority order for a plurality of communication terminals, and when there are a plurality of communication terminals that have determined that the decoded voice signal is sound. Designates one communication terminal having the highest priority among them, and transmits some of the speech coding parameters in the decoded speech signal of the designated communication terminal to each communication terminal without performing speech re-encoding processing. Therefore, even when speakers of a plurality of conference terminals speak simultaneously, there is an effect that the conference can be prevented from being confused.
[0094]
According to the present invention, when there are a plurality of communication terminals that determine that the decoded audio signal is sound, the voice encoding transmission system of the multipoint control device assigns the priority in the order of arrival of the audio signal received first. Since the setting is made, there is an effect that the progress of the conference is not disturbed by useless coughing, a hammering, or an interrupt message deviating from the rules.
[0095]
According to the present invention, the voice coded transmission system of the multipoint control device is configured such that one specific communication terminal is preferentially designated in advance and the priorities are set. There is an effect that even if a speaker competes by detecting a moderator's statement, a smooth conference can be performed.
[0096]
According to the present invention, some of the voice coding parameters in the voice coding and transmission system of the multipoint control device are configured to be parameters that carry pitch period information, so that the vocal cord frequency inherent to the voice of the speaker There is an effect that high-quality voice transmission can be realized by transmitting the information expressing the repetition cycle of this as it is.
[0097]
According to the present invention, since some of the speech encoding parameters in the speech encoding transmission system of the multipoint control device are configured to be parameters carrying spectrum envelope information, the vocal tract such as the oral cavity and nasal cavity of the speaker can be obtained. Is transmitted as it is, and high quality voice transmission can be realized.
[0098]
According to the present invention, when it is determined that the decoded audio signal received from the plurality of communication terminals is sound, the voice coded transmission system of the multipoint control device determines the number of the voice communication terminals according to the number of voice communication terminals. Since the configuration is such that the bit allocation of the frame of the voice coding parameter is set adaptively, there is an effect that the present invention can be applied to a transmission network where the transmission speed is constant or the transmission speed per channel is limited.
[0099]
According to the present invention, the voice coded transmission system of the multipoint control device is configured to transmit the decoded voice signal received from the communication terminal having the first priority and the communication terminal having the second priority among the plurality of communication terminals with sound. If it is determined that there are, two speech coding parameters respectively carrying the pitch cycle information and the spectrum envelope information in the decoded speech signal of the first communication terminal and the pitch cycle information in the decoded speech signal of the second communication terminal Are transmitted to each communication terminal without performing a voice re-encoding process, and a voice coding parameter for performing a voice re-coding process on the decoded voice signals of the first and second-rank communication terminals. Is transmitted to each communication terminal after performing degenerate quantization bit control based on the transmission rate control function, so that two speakers speak simultaneously. Even if there is an effect that it is possible to relatively good voice quality transmission.
[0100]
According to the present invention, when the voice coded transmission system of the multipoint control device determines that the decoded voice signal received from the main voice communication terminal and the sub voice communication terminal among the plurality of communication terminals is sound, There are two speech encoding parameters that respectively carry the pitch period information and the spectrum envelope information in the decoded speech signal of the main speech communication terminal, and the speech encoding parameter that carries the pitch period information in the decoded speech signal of the sub speech communication terminal. Is transmitted to each communication terminal without performing audio re-encoding processing, and the audio encoding parameters for performing audio re-encoding processing on the decoded audio signals of the main audio and sub audio communication terminals are degenerated based on the transmission rate control function. Is transmitted to each communication terminal after performing the quantization bit control of the main voice and the sub-voice simultaneously. Even if, by degenerating the coding parameters of the sub-audio, there is an effect that can be transmitted minimizing the degradation of the voice quality of the main voice while maintaining the transmission rate.
[0101]
According to the present invention, since the encoding parameter for performing the degenerate quantization bit control in the audio encoding transmission system of the multipoint control device is configured to be a gain codebook, it has an effect on the degradation of audio quality. Thus, there is an effect that even if two speakers speak at the same time, the deterioration of the voice quality can be suppressed even if the encoding parameters of the gain codebook index with less number are degenerated.
[0102]
According to the present invention, since the encoding parameter for performing the degenerate quantization bit control in the audio encoding transmission system of the multipoint control device is configured to be a noise codebook, it has an effect on the degradation of audio quality. In this case, the coding parameters of the algebraic codebook index with less number are degenerated, so that even if two speakers speak at the same time, there is an effect that the deterioration of the voice quality can be suppressed.
[0103]
According to the present invention, the voice coded transmission system of the multipoint control apparatus transmits the voice coded parameters to the respective communication terminals after performing the voice coded process or performs the voice coded process without performing the voice recoded process. Since the encoding mode information for deciding whether to transmit to the communication terminal is configured to be included in the frame of the encoding parameter consisting of a predetermined number of bits, it is possible to pass through a specific encoding parameter without affecting the transmission rate. There is an effect that the information of whether or not it can be included in the frame of the encoding parameter.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a multipoint control device applied to a speech coded transmission system according to Embodiment 1 of the present invention.
FIG. 2 is a block diagram showing a configuration of a speech coding parameter control unit according to the first embodiment.
FIG. 3 is a block diagram illustrating a configuration of an audio decoding processing unit according to the first embodiment.
FIG. 4 is a block diagram showing a configuration of a speech re-encoding processing unit according to the first embodiment.
FIG. 5 is a block diagram showing a configuration of an audio reproduction encoding processing unit according to the second embodiment.
FIG. 6 is a block diagram showing a configuration on a receiving side of the conference terminal according to the second embodiment.
FIG. 7 is a block diagram showing a configuration of an audio decoding processing unit of the conference terminal according to the second embodiment.
FIG. 8 is a diagram showing a frame configuration of speech coding parameters according to Embodiment 2.
FIG. 9 is an explanatory diagram showing an example of bit assignment of one frame of an encoding parameter according to Embodiment 2;
FIG. 10 is a block diagram showing a configuration of a multipoint control device according to the third embodiment.
FIG. 11 is a block diagram showing a configuration of a speech coding parameter control unit according to the third embodiment.
FIG. 12 is a block diagram showing a configuration of a multipoint control device applied to the speech coded transmission system according to Embodiment 4.
FIG. 13 is a block diagram showing a configuration of a speech coding parameter control unit according to the fourth embodiment.
FIG. 14 is a diagram showing a frame configuration of speech coding parameters according to Embodiment 5.
FIG. 15 is an explanatory diagram showing an example of bit assignment of one frame of an encoding parameter according to the fifth embodiment.
FIG. 16 is a block diagram illustrating a configuration of a conventional speech encoding device based on the CELP scheme.
FIG. 17 is a block diagram illustrating a configuration of a conventional multipoint control device.
FIG. 18 is a block diagram showing another configuration of a conventional multipoint control device.
FIG. 19 is a block diagram illustrating a configuration of a conventional CELP-based speech encoding device.
[Explanation of symbols]
104 linear prediction analysis processing unit, 105 LSP quantization processing unit, 106 LSP quantization codebook, 108 multiplexing unit, 109 inverse quantization processing unit, 110 adaptive codebook, 110b second adaptive codebook, 111 algebraic codebook (Noise codebook), 112, 116 adders, 113a, 113b gain control amplifier, 113c gain control amplifier, 114 gain quantization codebook, 115 synthesis filter, 117 auditory weighting filter, 118 distortion minimizing unit, 119 , 120 selector switch, 121 pitch pre-filter, 124 LPC / LSP converter, 125 gain MA predictor, 126 demultiplexer, 127, 128 control amplifier, 129 adder, 130 switch, 131a adaptive codebook, 131b gain Decoding section, 132 Decoding gain MA prediction section, 133 algebraic code decoding section, 134 pitch pref Filter, 135 LSP decoding section, 136 LSP interpolation section, 137 LSP / LPC conversion section, 138 synthesis filter, 139 post filter, 140 coding rate control section, 201 switch, 202 line interface section, 203 voice decoding processing section, 204 Voice detection section, 205 noise suppression processing section, 206 voice addition section, 207 distribution processing section, 208 own terminal voice subtraction section, 209 automatic gain control section, 210 voice recoding processing section, 211 voice coding parameter control section, 212 Selector, 213 distribution processing unit, 214 speaker selection unit, 215 first-arrival channel determination unit, 216
Priority speaker determination unit, 224 MCU control unit.

Claims (13)

A plurality of communication terminals are connected via a communication network, and coded voice signals obtained by coding voice handled by each communication terminal are to be received and transmitted, and predetermined processing is performed according to the information content of these coded voice signals. In the voice coded transmission system of the multipoint control device that distributes the coded voice signal processed to the plurality of communication terminals,
Decoding the encoded audio signal received from each communication terminal to generate a decoded audio signal,
Determining the communication terminal in which the decoded audio signal has sound,
Designate one of the communication terminals determined to be sound,
Of the plurality of types of speech encoding parameters in the decoded speech signal of the specified communication terminal, some of the speech encoding parameters are transmitted to the respective communication terminals without performing speech re-encoding processing,
The voice re-encoding process is performed on the plurality of types of voice encoding parameters in the decoded voice signal of the specified communication terminal and the plurality of types of voice coding parameters in the decoded voice signal of the other communication terminal out of the plurality of types of voice coding parameters. The voice coded transmission system of the multipoint control device, wherein the voice coded transmission system transmits the data to each of the communication terminals after the transmission. If only the decoded speech signal from one communication terminal is sound, the one communication terminal is designated, and speech re-encoding is performed for some speech encoding parameters in the decoded speech signal of the designated communication terminal. The voice coded transmission system for a multipoint control device according to claim 1, wherein the data is transmitted to each communication terminal without performing any processing. A priority is set for a plurality of communication terminals, and when there are a plurality of communication terminals that have determined that the decoded audio signal is sound, one communication terminal having the highest priority among the communication terminals is designated. 2. The voice code of the multipoint control device according to claim 1, wherein a part of voice coding parameters in the decoded voice signal of the designated communication terminal is transmitted to each communication terminal without performing voice re-encoding processing. Transmission system. 4. The voice of the multipoint control device according to claim 3, wherein when a plurality of communication terminals determine that the decoded voice signal is a sound, the priority is set in a first-come-first-served order in which the voice signal is received first. Coded transmission system. 4. The voice coded transmission system for a multipoint control device according to claim 3, wherein one specific communication terminal is specified in advance and priorities are set. The speech encoding transmission system for a multipoint control device according to any one of claims 1 to 5, wherein some speech encoding parameters are parameters carrying pitch period information. The speech encoding transmission system of the multipoint control device according to claim 1, wherein some speech encoding parameters are parameters carrying spectrum envelope information. If the decoded speech signals received from the plurality of communication terminals are determined to be speech, the bit allocation of the speech coding parameter frame is adaptively set according to the number of speech communication terminals. The voice coded transmission system for a multipoint control device according to any one of claims 1 to 7, wherein: When it is determined that the decoded voice signals received from the first and second communication terminals are sound, the decoded voice of the first communication terminal is determined. The speech re-encoding process is performed on two speech encoding parameters respectively carrying the pitch period information and the spectrum envelope information in the signal and the speech encoding parameter carrying the pitch period information in the decoded speech signal of the second communication terminal. For each of the speech encoding parameters to be transmitted to each communication terminal and to perform speech re-encoding processing on the decoded speech signals of the first and second communication terminals, degenerate quantization bit control based on a transmission rate control function is performed. 9. The voice coded transmission system for a multipoint control device according to claim 8, wherein the transmission is performed to each of the communication terminals after performing. When it is determined that the decoded voice signal received from the communication terminal of the main voice and the communication terminal of the sub-voice among the plurality of communication terminals is sound, the pitch period information and the pitch period information in the decoded voice signal of the communication terminal of the main voice are determined. Two voice coding parameters each carrying spectrum envelope information and a voice coding parameter carrying pitch period information in the decoded voice signal of the sub-voice communication terminal are transmitted to each communication terminal without performing voice re-encoding processing. The voice coding parameters for performing voice re-encoding on the decoded voice signals of the communication terminals of the main voice and the sub-voice are transmitted to the respective communication terminals after performing degenerate quantization bit control based on a transmission rate control function. 9. The voice coded transmission system for a multipoint control device according to claim 8, wherein: 11. The speech encoding transmission system for a multipoint control device according to claim 9, wherein the encoding parameter for which the quantization bit control of the degeneration is performed is a gain codebook. 11. The speech encoding transmission system for a multipoint control device according to claim 9, wherein the encoding parameter for which the degenerate quantization bit control is performed is a noise codebook. A predetermined bit is set as coding mode information for determining whether to transmit to each communication terminal after performing a voice re-encoding process on a voice coding parameter or to transmit to each of the communication terminals without performing a voice re-coding process. The speech encoding transmission system for a multipoint control device according to any one of claims 1 to 12, wherein the parameter is included in a frame of encoding parameters consisting of numbers.

Images