JP2002026964A - Voice packet transmitter, voice packet receiver, and packet communication system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice packet receiver that minimizes deterioration in voice quality. SOLUTION: The voice packet receiver that receives a voice packet sent from a voice packet transmitter via a prescribed network, is provided with a reception state detection means that detects a reception state of a voice packet fluctuated corresponding to the traffic on the network and with a reception state information transmission means that transmits the reception state information in response to the reception state detected by the reception state detection means to the network to transmit the reception state information to the voice packet transmitter.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a packet communication system, and can be applied, for example, to voice communication using a packet communication network such as the Internet.

[0002] The present invention also relates to a voice packet receiving apparatus as a component of such a packet communication system.

Further, the present invention relates to a voice packet transmitting apparatus as a component of such a packet communication system.

[0004]

2. Description of the Related Art At present, a method of performing voice communication using a packet communication network such as the Internet (a so-called Internet telephone) is active.

[0005]

However, in order to ensure good sound quality in such packet communication,
The relationship between bit rate and sound quality is an important issue.

[0006] Increasing the bit rate reduces coding distortion and improves sound quality, but increases traffic on the network, so that packet arrival time becomes unstable, and buffering delay and packet loss (packet loss) are tolerated. Inevitably, the sound quality deteriorates as a result. On the other hand, if the bit rate is reduced, the problem of traffic is alleviated, but the coding distortion increases, and the sound quality also deteriorates.

One of the solutions to this problem is
A silence compression method is used. The silence compression method is a method used to reduce the number of transmission bits during a silence period of voice. The transmitting side determines, for each frame, whether the frame is speech or silence (hereinafter, “speech decision”). In the case of a silent frame, this method encodes the information into a smaller amount of information than a voiced frame and sends the information to the network. In human-to-human conversations, generally 50% to 60% per speaker
Is said to be silent, and the silent compression method is an extremely effective coding method for lowering the average bit rate.

However, it is technically very difficult to determine whether or not a sound is to be determined in a specific communication device.
If it is set too high, a frame that should be an utterance section is judged as silence. Conversely, if the sound judgment level is set too low (ie, low), even a frame in a section where no utterance is made In some cases, it may be determined that there is sound, and this determination error leads to sound quality degradation. Therefore, if it is relaxed from the viewpoint of the amount of traffic on the network, it is preferable in terms of sound quality not to perform silent compression as much as possible.

FIG. 2 shows the relationship between the average bit rate and the sound quality when the silence compression level Thresh (the sound determination level) is changed. The horizontal axis represents the average bit rate without silence compression (ie, when Thresh = 0) as 1
The average bit rate is shown as 00%, and the vertical axis indicates the sound quality evaluation by the MOS test (the higher the numerical value in the five-step evaluation, the higher the sound quality).

It can be seen that as the threshold is increased, the bit rate decreases and the sound quality also decreases. Therefore, in order to ensure sound quality, it is understood that it is preferable to use a low threshold as long as no problem occurs in the network at the time of transmission. Since the traffic on the network changes with time, it is preferable that the Thresh is also dynamically changed and adapted to the fluctuation of the traffic.

As a method of dynamically changing the sound determination level, a VAD (voiced detection) algorithm that sequentially adapts to the noise level is known.

The following document 1 describes the VAD algorithm.

Reference 1: ITU-T Recommendation G. 723.1
Annex A However, the method of Document 1 aims at adapting the sound determination level to the background noise that changes over time, and is independent of the traffic volume on the network.

The following document 2 describes a method of adapting the transmission data rate according to the amount of traffic on the network.

Reference 2: Japanese Patent Application Laid-Open No. H11-177623 However, the method of Reference 2 adapts a bit rate to an accurate value specified by using a multi-rate audio encoder, and specifies a sound determination level. Not something.

After all, according to the methods described in Documents 1 and 2, the sound detection level is changed in a fixed time unit within a range that does not impose a load on the traffic volume on the network, and the sound quality is deteriorated. The task of providing packet voice communications with a minimum has not been achieved.

[0017]

According to a first aspect of the present invention, there is provided a voice packet receiving apparatus for receiving a voice packet transmitted from a voice packet transmitting apparatus via a predetermined network. Receiving status detecting means for detecting a receiving status of a voice packet that fluctuates according to traffic on the network, and receiving status information according to the receiving status detected by the receiving status detecting unit;
And a receiving status information transmitting means for transmitting to the network for transmitting to the voice packet transmitting apparatus.

According to a second aspect of the present invention, in a voice packet transmitting apparatus for transmitting a voice packet to a voice packet receiving apparatus via a predetermined network, (1) the voice frame to be transmitted is voiced or silent. Determining means for determining whether or not there is a sound frame in accordance with the determination level; and (2) encoding a silent frame determined to be silent by the determining means so as to have a smaller amount of information than a sound frame determined to be sound. Voice encoding means, and (3) receiving, via the network, reception status information transmitted by the voice packet receiving apparatus in accordance with a reception status of the voice packet, which varies according to traffic on the network. Receiving status information receiving means; and (4) a determination level changing means for changing the determination level in accordance with the receiving status information received by the receiving status information receiving means. Characterized in that it comprises a means.

Further, a packet communication system according to a third aspect of the present invention includes the voice packet receiving device of claim 1 and the voice packet transmitting device of claim 2.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (A) Embodiment Hereinafter, embodiments of a voice packet transmitting device, a voice packet receiving device, and a packet communication system according to the present invention will be described.

(A-1) Configuration of the First Embodiment FIG. 1 shows the overall configuration of a communication system 10 of the present embodiment.

In FIG. 1, the communication system 10 includes:
A voice transmitting terminal 11, an IP (Internet Protocol) network 12, and a voice receiving terminal 13 are provided.

The voice transmitting terminal 11 includes a voice input unit 20, a voice encoding unit 21, and a voice packet transmitting unit 22.
, A reception state information receiving section 23, and a sound determination level calculation section 24.

The audio input section 20 is a section which takes in an externally input audio by, for example, a microphone or the like, converts the audio into audio data AD for each frame length, and outputs the audio data AD.

The audio encoding unit 21 which receives the audio data AD from the audio input unit 20 converts the audio data AD into an audio frame (ie, audio frame encoded data) CF for each fixed time length (for example, 30 ms). This is the part that converts and outputs. Since the speech encoding unit 21 is a part that executes silence compression, the speech frame CF to be output is a silence frame (silence frame encoded data).
Either CF2 (see FIG. 3) or a voiced frame (voiced frame coded data) CF1.

That is, the speech encoding unit 21 has an internal configuration as shown in FIG. 3, and includes a speech determination unit 21A, a speech frame encoding unit 21B, and a silence frame encoding unit 21C.

The sound determination unit 21A executes the sound determination in accordance with a sound determination level (voice detection level) LTH supplied from a sound determination level calculation unit 24, which will be described later, and outputs one frame of voice. For each data AD, the audio data AD determined to be voiced is supplied to the voiced frame encoding unit 21B, and the voiced data AD determined to be silence is supplied to the silent frame encoding unit 2B.
This is the part that supplies 1C.

In sound determination, the sound determination unit 21A
The variance of the audio data AD for one frame is calculated. If the variance is equal to or higher than the sound determination level LTH, it is determined that there is sound, and if the value is less than or equal to the sound determination level LTH, there is no sound.

The voiced frame coding unit 21B executes the coding in which the coded data size is relatively large for the voiced frame and outputs the voiced frame CF1, and the voiceless frame coding unit 21C outputs the voiceless frame CF1. This is a portion for executing the encoding for reducing the encoded data size and outputting the silent frame CF2. For example, the voiced frame encoding unit 21B uses a 16-bit PCM (Pulse Code M
odulation), and the silence frame coding 21C may be coded by a 4-bit PCM method.

The voice packet transmitting section 22 shown in FIG. 1 is a section for forming a voice packet AP based on the voice frame CF and transmitting it to the network 12. One voice packet AP may contain one or a plurality of voice frames CF, but usually, a plurality of frames (for example, five frames)
It is housed in P.

The voice packet transmitting section 22 includes a continuous information providing section 22A for providing continuous information indicating the continuous state of the voice packet sequence to the packet header of each voice packet AP transmitted in time series. . In the present embodiment, this continuous information is assumed to be a packet number (pID) that is incremented by one according to the transmission order.

The receiving state information receiving section 23 is a part for receiving the receiving state information SI from the network 12. Since the network 12 is a packet communication network, the reception state information SI is also accommodated in the packet RP and transmitted. The reception state information receiving unit 23
The reception state information SI is extracted from P and output to the sound determination level calculation unit 24.

The sound level judgment section 24 is a section for changing the sound level LTH according to the reception status information SI.

The sound level LTH is a threshold value corresponding to the silence compression level Thresh. Therefore, as described above, if the sound determination level LTH is set too high, the audio data A of the frame that should be the utterance section is originally set.
D is determined to be silent, and conversely, the sound determination level L
If the TH is set too low, the audio data AD of a frame in a section in which nothing is uttered may be determined as having sound, leading to sound quality deterioration.

The sound determination level calculation unit 24 minimizes such sound quality deterioration by adapting the sound determination level LTH to the reception state information SI.

On the other hand, the voice receiving terminal 13 which receives the voice packet AP via the network 12 and also transmits the packet RP includes a voice packet receiving unit 30, a voice decoding unit 31, a voice output unit 32 And a reception state information creation unit 33 and a reception state information transmission unit 34.

The voice packet receiving section 30 receives the voice packet AP transmitted from the voice packet transmitting section 22 in the voice transmitting terminal 11 from the network 12. The voice packet receiving unit 30 supplies the voice frame CF extracted from the received voice packet AP to the voice decoding unit 31, and supplies the packet number pID to the reception state information creation unit 33 as reception state data SD indicating the reception state. I do.

The audio decoding unit 31 that has received the audio frame CF converts the audio frame CF into audio data AD and outputs it to the audio output unit 32.

The audio output section 32 is a section for outputting an audio corresponding to the audio data AD, for example, by a speaker or the like.

The receiving state information creating section 33 is a section that creates receiving state information SI by performing processing described later based on the receiving state data SD, and outputs it to the receiving state information transmitting section 34.

The reception state information transmitting section 34 creates a packet RP containing the reception state information SI and sends it to the network 12. The packet RP is provided with routing information designating the voice transmitting terminal 11 as a destination.

The voice transmitting terminal 11 and the voice receiving terminal 13 can be realized by, for example, a PC (personal computer) having an IP network interface and a sound input / output unit.

The operation of this embodiment having the above configuration will be described below.

(A-2) Operation of the First Embodiment An audio signal input from outside is converted into audio data AD in units of one frame by an audio input unit 20 and supplied to an audio encoding unit 21.

The voice packet AP is transmitted to the network 12 after being encoded by the audio encoding unit 21 and packetized by the audio packet transmitting unit 22.

The voice packet AP includes a voice receiving terminal 1
Since the routing information specifying 3 is given,
The data is transferred to the voice receiving terminal 13 by the IP network 12.

When the voice transmitting terminal 11 successively sends out voice packets AP in time series, the continuous information providing unit 22A
Assigns “1” (decimal notation) to the first audio packet AP (1) as the packet number pID.

Similarly, the second voice packet AP
(2) has a packet number pID of “2”, a third voice packet AP (3) has a packet number pID of “3”, and a fourth voice packet AP (4) has a packet number pID. ,..., And the Nth audio packet AP (N) is assigned “N” as the packet number pID.

Therefore, if the traffic of the network 12 is within the normal range and the degree of congestion is normal, the reception state at the voice receiving terminal 13 becomes normal, and the voice packets AP (1) to AP (N) become 1 Packets are also received in this order without dropping. At this time, the reception state data SD output by the voice packet receiving unit 30 in the voice receiving terminal 13 is a packet number pID of continuous “1” to “N” without omission which is incremented by one.

However, if the traffic of the network 12 exceeds the normal range and the congestion degree becomes higher than normal, a packet loss occurs in the network 12 and the voice packets AP (1) to AP (N) are received. Something cannot be done.

At this time, the voice packet receiving unit 30
Is missing in the reception state data SD output by the.

When the reception state data SD is, for example, 1, 2, 2,
If 3, 4,..., M−1, M + 1,.
This is the case where the third voice packet AP (M) is lost due to packet loss.

The conversion of the reception state data SD (received packet number pID) into the reception state information SI is performed by the reception state information creation unit 33 as shown in FIG.
It depends on the processing.

The flow chart of FIG.
It consists of each step. In FIG. 4, the number of packet losses per 100 packets (that is, the packet loss rate) is calculated and output as the reception state information SI.

In FIG. 4, when the process of the reception state information creating unit 33 starts, "0" is substituted into the variables Num, Loss, and Last_pID as initial values (S10). Here, the variable (packet counter) Num indicates the number of received voice packets AP, the variable (packet loss counter) Loss indicates the number of voice packets AP lost due to packet loss, and the variable Last_pID will Voice packet AP (X) to be processed using the flowchart of FIG.
Indicates the pID held by the voice packet AP (X-1) received immediately before.

Note that a variable means an area secured on a memory for the purpose in terms of hardware.
The same applies to the following.

Next, the function Get_pID acquires the value of the pID of the audio packet AP (X) supplied from the audio packet receiving unit 30 as the reception state data SD, and substitutes that value for the variable pID (S11). This function Get_p
The ID performs the same process each time a new packet number pID is supplied from the voice packet receiving unit 30.

In step S12, it is checked whether the value of the variable Num is equal to "100" (decimal notation). If the value is equal, the flow branches to the Yes side, and if not, the flow branches to the No side.

First, since the above-mentioned "0" is substituted for the variable Num, the No side is selected for this branch. Even in the earliest case, step S12 branches to the Yes side when the 101st voice packet AP (101) is received. If there is a packet loss, this 101
Becomes larger by the packet loss number, and the step S1
The timing at which the branch on the Yes side of No. 2 is selected is delayed.

X (1 ≦ X ≦ 101, where X is a natural number)
When processing the audio packet AP (X) which is the third packet, the value of the variable Num is X−1 if there is no packet loss until then.

When step S12 branches to No, the processing of step S13 is performed.

In step S13, the voice packet A
It is checked whether the value of the variable pID of P (X) is equal to the value of the Last_pID plus "1". That is, it is checked whether the pID supplied from the voice packet receiving unit 30 as the reception state data SD is incremented one by one and there is no omission. If there is no omission, step S13 branches to Yes side, and the process proceeds to step S17. If there is omission, the process branches to No side and proceeds to step S16.

For example, since the initial value of Last_pID is “0”, the voice packet AP (1) transmitted from the voice transmitting terminal 11 first is received by the voice receiving terminal 13 without being lost in the network 12. Then, the pID “1” of the voice packet AP (1) and La
The values of st_pID + 1 are equal.

When processing the Xth voice packet AP (X), the voice packet AP (X) or AP (X-
Only when both of 1) are not lost due to the packet loss, the Yes side of step S13 is selected.

In step S17, which is performed after step S13, Last_pID ++ and Num
++ is performed. That is, the variable Last_pID and the variable Num are incremented. Step S17
Then, the step S11 is executed.

On the other hand, in step S16 which is performed after step S13, the variables Last_pID and Nu
In addition to the variable m, the variable Loss is also incremented, and the process returns to step S12.

When the step S12 branches to the Yes side, at step S14 Send (Loss)
The processing of s) is performed. That is, the transmission function Send ()
As a result, the value of the variable Loss is output to the reception state information transmission unit 34 as the reception state information SI.

After step S14, in step S15, the initial value "0" is substituted for the variable Loss and the variable Num, and the process proceeds to step S13. At this time, since the variable Last_pID is not initialized, the relation between the variable pID and Last_pID in step S13 is maintained. That is, as the packet number pID, a serial number of 101 or more can be used.

As long as normal communication without packet loss is maintained between the voice transmitting terminal 11 and the voice receiving terminal 13, the loop composed of steps S11, S12, S13, and S17 is repeated, and the loop 101
The processing of steps S14 and S15 is performed for each repetition (every 100 packet periods). In this case, since the packet loss number is “0”, the function Send (Loss)
Is "0".

However, if there is a packet loss, the loop constituted by steps S12, S13 and S16 is repeated by the number of the packet loss.

At step S 14, the reception state information transmitting section 34, which has received the value of the variable Loss as the reception state information SI, stores the value in the packet RP and sends it to the voice transmission terminal 11 to the network 12. Send out.

In the voice transmitting terminal 11 that has received the packet RP from the network 12, the reception state information receiving unit 23 extracts the value of the variable Loss as the reception state information SI from the packet RP, and calculates the sound determination level. Output to the unit 24.

The sound level judgment section 24 calculates the sound level LTH based on the value of the variable Loss.
The process of setting Threshold in the audio encoding unit 21 is shown in the flowchart of FIG.

The flowchart of FIG.
It consists of each step.

In FIG. 5, the sound determination level calculating section 24
Is started (S20), the function Get_Lo
ss is the reception state information receiving unit 2 as the reception state information SI.
The value of the variable Loss supplied from No. 3 is obtained and substituted into the variable Loss on the sound determination level calculation unit 24 side (S21). The function Get_Loss repeats the same operation each time new reception state information SI is supplied.

In the flowchart of FIG. 5, the variable Loss means the variable Loss on the sound determination level calculation unit 24 side. This variable Loss also
The variable Loss on the reception state information creation unit 33 side is the same in that the packet loss number is stored, but the variable Loss on the reception state information creation unit 33 side is one 100.
While the packet loss counter is incremented during the packet period and its value can change, the value of this variable Loss is constant during one 100 packet period.

Next, in step S22, the function Get_T
hresh is Thresh as a value of the sound determination level LTH used in the voice encoding unit 21 in the immediately preceding 100 packet period (may be the voice determination level LTH currently used in the voice encoding unit 21). Is obtained and assigned to the variable Threshold0. Function Get_Thresh
The same operation is repeated every time new reception state information SI is supplied.

Here, the variable Threshold0 is set to the immediately preceding 1
This is a variable for storing the value of the sound determination level TLH used by the voice encoding unit 21 during the 00 packet period.

The value of Threshold is a value to be compared with the variance of the audio data AD for one frame by the sound existence determination unit 21A. The larger the value is, the higher the sound determination level LTH is.
The tendency that the determination result that the audio data AD is a silent frame increases, and conversely, the smaller the value is, the lower the voiced determination level LTH is, and the determination result that the voice data AD is a voiced frame is increased. Increased tendency to be issued.

In step S23 following step S22, it is checked whether or not the value of the variable Loss is "0". If the value is "0", the process branches to the Yes side and the process proceeds to step S24, in which "0" is set. If not, the flow branches to No and proceeds to step S27.

In step S24, the variable Thres
A value obtained by subtracting a predetermined change width α (α> 0) from the value of h0 is substituted for a variable Thresh. here,
The variable Threshold is basically a sound determination level LTH used by the voice encoding unit 21 in the next 100 packet periods.
Is a variable for storing.

In this processing, if the packet loss number during the immediately preceding 100 packet period is “0”, the bit rate is considered to be sufficiently low at present and there is no traffic problem, and the coding rate is increased by increasing the bit rate. That is, the control is performed in the direction of decreasing.

In step S25 executed after step S24, the value of the function Threshold is limited by the function Limitation within an effective range of a predetermined maximum value or less or a minimum value or more. That is, if the value of the variable Thresh is within the valid range, the value is used as it is, and if the value is larger than the valid range, the maximum value defined by the function Limit is set as the value of the variable Thresh. If the value is smaller than the range, the minimum value defined by the function Limitation is set as the value of the variable Threshold.

Next, at step S26, the variable Thr
The value of esh is set as a sound determination level LTH of the audio encoding unit 21 by the function SetEnc, and the processing is performed as follows.
The process returns to step S21.

On the other hand, in step S27, which is executed when the branch on the No side is selected in step S23, a value obtained by adding Loss * β to the value of variable Thresh0 is substituted for the variable Thresh. Here, “*” is an operator meaning multiplication, and Loss * β
Means a value obtained by multiplying the value of the variable Loss by β (> 0).

In this processing, if the packet loss number during the immediately preceding 100 packet period is not “0”, the bit rate is still too high at present and the traffic problem is large, even if the coding distortion increases. This means that the bit rate is controlled to decrease.

The value of β may be equal to or different from the value of α.

After step S27, steps S25 and S26 are also executed.

Then, the audio encoding unit 21 determines in step S
Variable Th set as sound determination level LTH in 26
Based on the value of “resh”, a sound determination is made for the next 100 packet periods.

By repeating the above procedure, the last 100
Voice communication can be performed at the sound determination level LTH reflecting the packet reception state.

In FIG. 5, when the traffic of the network 12 is stable, the Yes side and the No side in step S23 are selected with almost equal frequency, and the sound determination level LTH is over a certain value (optimum value). , The state of fluctuating with a small amplitude is maintained.

If the traffic of the network 12 fluctuates, the selection frequency of the Yes side and the No side in step S23 becomes uneven, and follows the fluctuation of the optimum value.

In short, in the present embodiment, the reception state of the voice packet AP is calculated, and when the reception state of the voice packet AP is poor, the sound determination level LTH is increased, and when the reception state is clear, the sound determination level LTH is raised. Can be lowered.

Thus, when traffic on the network 12 is congested, the average bit rate of the voice packet AP is reduced to suppress sound quality deterioration due to packet loss and increase in jitter. Sound quality degradation due to compression can be suppressed, and the average bit rate can be automatically adapted to network congestion.

The flowchart of FIG. 4 (particularly, a loop constituted by steps S11, S12, S13 and S17, or steps S11, S12, S13 and S1)
6 is normally repeated 100 times in 100 packet periods. However, the flowchart of FIG. 5 shows that the step S is performed in 100 packet periods or in a period between 100 packet periods and 100 packet periods.
Loop formed by 21, S22, S23, S24, S25, S26 or steps S21, S22, S
The loop constituted by 23, S27, S25, and S26 may be executed only once.

(A-3) Effects of the First Embodiment As described above, according to the present embodiment, by setting an appropriate average bit rate corresponding to the congestion state of the network (12), a high average bit rate can be obtained. In addition to preventing packet loss caused by an excessively low bit rate, it also minimizes audio quality degradation due to coding distortion caused by an excessively low bit rate. Highly reliable voice communication can be realized.

In a communication state in which packet loss occurs, it is highly possible that jitter increases with packet loss. However, if packet loss is prevented in the present embodiment, it is possible to reduce deterioration in sound quality due to the jitter. Is possible.

(B) Second Embodiment In the following, only differences between the present embodiment and the first embodiment will be described.

In the present embodiment, a silence compression method in which the sound determination level LTH dynamically changes according to the content of the audio data AD is applied inside the audio encoding unit.

This embodiment is characterized in a portion related to the variable Scale calculated by the sound determination level calculation unit.

Further, the present embodiment can realize the speech coding method described in the above-mentioned reference 1.

(B-1) Configuration and Operation of Second Embodiment The overall configuration of a communication system 40 of the present embodiment is the same as the communication system 10 of the first embodiment shown in FIG.
Therefore, the same reference numerals 20, 22, 2 as in the first embodiment.
The functions of the units denoted by 3, 30 to 34 are the same as those of the first embodiment.

However, unlike the sound judgment level calculation section 24, the sound judgment level calculation section 44 of the present embodiment receives the reception state information SI (Loss) and sets the sound judgment level LTH (Thresh) instead of the sound judgment level LTH (Thresh). Sound judgment scale S
This is the part for calculating the call. Based on the reception status information SI, the sound determination level calculation unit 44
The process of obtaining the call is as shown in FIG. 7 described later.

Also, the internal configuration of the speech encoding unit 41 is different from that of the speech encoding unit 21 of the first embodiment, as shown in FIG. The speech encoding unit 41 according to the present embodiment includes an ACELP (Algebr
aic Code Exited Liner Prediction).

The principle of the ACELP system incorporating the VAD system is as described in detail in the above-mentioned reference 1, but in a normal VAD system as described in the reference 1, the LPC (Linear Predictive Coding) is used. ) Although the sound determination level Thresh (Nlev) based on the average residual power Err and the noise level Nlev by the analysis is determined as a threshold, in the present embodiment, the threshold is determined in the present embodiment.
Is set to the value of S read from the sound determination Scale accumulation unit 45.
Scale multiplied by the value of call (initial value is 1.00)
* Thresh (Nlev) is set as threshold SC.

Therefore, in the method of Reference 1, it is impossible to reflect the packet loss rate (the reception state SI) on the threshold, but in the present embodiment, it is possible.

Further, when compared with the first embodiment, the first embodiment
The threshold (voice determination level LTH) of the embodiment is generated according to the flowchart of FIG. 5, whereas the threshold of the embodiment is generated based on the average residual power Err and the noise level Nlev. The difference is that the threshold value is Thresh (Nlev).

In FIG. 6, the speech encoding unit 41
It includes a voiced frame coding unit 41B, a voiceless frame coding unit 41C, a voice analysis unit 42, a voiced determination unit 43, and a voiced determination Scale storage unit 45.

The voiced frame coding unit 41B corresponds to the voiced frame coding unit 21B, the voiceless frame coding unit 41C corresponds to the voiceless frame coding unit 21C, and the voiced sound determination unit 43 Corresponds to the sound presence determination unit 21A.

However, the sound determination section 43 is the same as the sound determination section 21A in that the sound determination is performed, but the voice determination section 43A or the silent frame coding section 41A performs the voice determination.
The function of supplying the audio data AD to C is provided with the audio analysis unit 42, and also has various differences described later in other points.

The voice analysis unit 42 that receives the audio data AD for one frame from the voice input unit 20 performs a voice analysis on the voice data AD, and supplies the analysis result to the sound determination unit 43.

In the present embodiment, as the voice analysis, L
A PC analysis is performed, and the analysis result and incidental information (for example, the power calculation result and the number of zero crossings) that can be extracted from the audio data AD are stored in the Thresh (Nl
ev) is supplied to the sound existence determination unit 43 as basic information BI used for generating ev).

[0113] The sound existence judgment unit 43 has a function of generating a Threshold (Nlev) based on the basic information BI.

The basic information BI is defined as Thresh (N
lev) is generated, and also used as an object of sound determination.

On the other hand, the sound determination Scale storage unit 45
From the sound judgment level calculation unit 44, the sound judgment scale Sca is calculated.
le is received and stored, and is supplied to the sound existence determination unit 43.

The sound existence judging section 43 performs the above-mentioned Thresh.
(Nlev) is multiplied by this sound determination scale Scale to obtain a determination threshold SC (= Scale * Thresh)
(Nlev)). This threshold SC for determination is a sound determination level used for sound determination in the sound determination unit 43.

When the judgment result JA of the sound judgment by the sound judgment section 43 is supplied to the sound analysis section 42, the sound analysis section 42 outputs the sound data AD for one frame in accordance with the judgment result JA. It is supplied to either the voiced frame coding unit 41B or the silent frame coding unit 41C.

Next, the processing of the sound judgment level calculation unit 44 for supplying the sound judgment scale Scale to the sound judgment scale accumulation unit 44 will be described. This processing is shown in FIG.
As shown in the flowchart of FIG.
oss) is converted to a sound determination scale Scale.

The flowchart of FIG.
It consists of each step.

In FIG. 7, when the process starts (S
30), the real number variable Scale stores a real number 1.00 as an initial value.

Next, the function Get_Loss acquires the value of the variable Loss supplied from the reception state information receiving section 23 as the reception state information SI, and substitutes it for the variable Loss on the sound determination level calculation section 24 side ( S31).

This step S31 is performed in step S2.
This is exactly the same processing as 1. Therefore, step S3
Even if it is 1, the function Get_Loss repeats the same operation every time new reception state information SI is supplied.

If the value stored in the variable Loss is “0”, the process proceeds to step S33,
If not "0", the process proceeds to step S35.

In step S33, the immediately preceding variable Scal
The value (Scale * α) obtained by multiplying the value of e (the initial value is the initial value 1.00) by α is substituted for the variable Scale. Here, α is a real constant of 0 <α <1.

In step S34 following step S33, the value of the variable Scale is set as the determination threshold SC of the soundness determination unit 43 by the function SetEnc.
The process returns to step S31.

On the other hand, in step S35, the result of multiplying the value of the variable Scale immediately before by the value of the variable Loss and the value of the constant β is assigned to the variable Scale. Here, the constant β is a real number (or an integer) of 1 <β.

After the step S35, the step S34 is performed, and the value of the variable Scale is changed to the function Set.
The data is supplied to the sound determination unit 43 by Enc, and the process returns to the step S31.

According to the flowchart of FIG. 7, if the state where there is no packet loss continues, the constant α is raised to the power and the value of the variable Scale decreases.
On the other hand, when the state with the packet loss continues, the constant β is raised to the power and the determination threshold SC increases.

If the traffic on the network 12 is stable and the contents of the audio data AD are also stable, the Yes side and the No side in step S32 are selected with almost equal frequency, and the determination threshold SC is set to a fixed value (optimal value). ) To maintain a state of fluctuation with a small amplitude.

If the traffic of the network 12 or the content of the audio data AD fluctuates, the selection frequency of the Yes side and the No side in step S32 becomes unequal, and follows the fluctuation of the optimum value.

In the first embodiment, the presence / absence level LTH changes only when the traffic on the network 12 fluctuates and the reception state information SI changes. In the present embodiment, the traffic on the network 12 is substantially constant. However, if the content of the audio data AD changes, the determination threshold SC may change according to the change.

(B-2) Effects of the Second Embodiment According to the present embodiment, the same effects as those of the first embodiment can be obtained.

In addition, in the present embodiment, the judgment threshold (S
C) dynamically fluctuates according to the content of the audio data (AD) supplied from the audio input unit (20), so that more appropriate and precise control of the bit rate is possible.

(C) Other Embodiments In the first and second embodiments, the case where the voice packet AP is transmitted and received between the two terminals 12 and 13 has been described. As shown in FIG. Alternatively, a system configuration in which a large number of terminals 51 to 56 are connected to the same network 12 to transmit and receive voice packets between a plurality of sets of terminals may be adopted.
In the case of FIG. 8, the terminals that communicate with each other, such as the terminals 11 and 13, must be compatible with the present invention. However, terminals that do not support the present invention may be mixed in the system. Absent.

In the first and second embodiments, a PC (personal computer) having an IP network interface unit and a sound input / output unit has been described as an example of realizing the voice transmitting terminal 11 and the voice receiving terminal 13. However, each PC can be replaced with a combination of a plurality of devices as shown in FIG.

For example, a combination of a telephone 64, a PBX (Private Branch Exchange) 63 and a telephone-packet converter 62, a combination of a telephone-packet converter 60 compatible with the IP protocol and a telephone 61, etc. Each PC may be replaced. As a conversion device, for example, V
An oIP (voice over IP) gateway device such as BS-1200 manufactured by Oki Electric Corporation can be used.

Although the packet number pID is used as the continuous information in the first and second embodiments, for example, time stamp information or the like can be used instead of the packet number pID. The time stamp information is information indicating the time at which the voice packet AP was transmitted or the time at which the voice packet AP was generated.

The outline of the processing when a time stamp is used as continuous information will be described below with reference to FIG.

First, the continuation information adding unit 22A in the voice packet transmitting unit 22 adds a time stamp to each voice packet AP when transmitting each voice packet AP.

Next, the voice packet receiving unit 30 compares the time stamp of each received voice packet AP with the current time, and calculates the time required for transferring the voice packet AP (packet transfer time).

The reception state information creating section 33 obtains a packet fluctuation time (jitter) by comparing a predetermined average packet transfer time with a packet transfer time of each voice packet AP, and obtains a jitter dispersion every 100 packets. Is calculated, and this is used as reception state information SI.

The method of adding continuous information such as the packet number pID and the time stamp information to the voice packet AP may be any method, and may be RTP (RealTiP)
An existing protocol such as meTransport Protocol may be used.

The transmission interval of the reception status information, that is, the transmission interval of the packet RP is arbitrary, but if it is performed too frequently, the transmission of the packet RP itself becomes a load on the traffic of the network (12). is necessary.

Further, as the encoding method, PCM
Although the above description has been made using 16 bits and 4 bits, any method may be used as long as the amount of encoded audio data per frame is smaller in a silent frame than in a voiced frame. For example, ITU-T Recommendation G. G.729 Annex B. 729 may be used.

In the first and second embodiments, the method of dynamically controlling the sound determination level LTH has been described. However, the control may be performed by binary control for turning on / off the silent compression function. In this case, the present invention can be realized without making any change to the audio encoding unit having such an ON / OFF switch.

In the first and second embodiments, with respect to the voice packet AP, the voice transmitting terminal 11 has only the transmitting function, and the voice receiving terminal 13 has only the receiving function. , 13 may be provided with both a voice packet transmitting function and a voice packet receiving function to enable bidirectional voice packet communication.

In this case, the reception status information SI can be contained not in an independent packet RP but in a part of a voice packet transmitted from the terminal 13 to the terminal 11. .

In this case, as shown in FIG. 10, a sound determination level calculation unit 81 may be provided in the reception terminal (transmission / reception terminal) 71.

Further, in step S12 of FIG. 4, to check the number of packet losses per 100 packets of the voice packet AP, it was checked whether or not the value of the variable Num was equal to "100". Can be changed to an arbitrary constant other than 100, and the number of packet losses for each arbitrary number of packets can be inspected.

As the reception status information, instead of the reception status information SI, other characteristic amounts (for example, the degree of jitter and the degree of delay time) of the loss rate may be inspected. That is, as the reception state information, the voice packet AP related to the traffic volume on the network
Any variable may be used as long as it is a variable that indicates the reception state of.

In the flowchart of FIG. 5, the constants α and β are used to change the sound determination level LTH. However, when it is determined that the packet reception state is bad, the sound determination level LTH is increased and the packet reception level is increased. If it is determined that there is enough space, other methods can be used as long as the method can lower the sound determination level LTH.

For example, the reception state of the past several periods may be stored, and the level of the sound determination level may be changed in a manner reflecting the reception state. As an example, Loss for 10 consecutive periods
For the first time when = 0, a method of increasing the sound determination level LTH by α may be used.

In the flowchart of FIG. 5, the variable L
Two types of loops are executed depending on whether the value of oss is “0”. However, even in the case where the value of oss is not “0”, the value range of the variable Loss can have a width of 1 to 100. For example, different loops may be executed for “1” and “10”.

Further, since the flow chart of FIG. 4 has a very large number of repetitions as described above, the processing of this flow chart may be a bottleneck in improving the operation speed of the entire system. Therefore, if the flow chart of FIG. 4 is configured more efficiently, there is a great advantage that the speed of the entire system can be increased.

For example, depending on conditions such as the degree of congestion of the network (12), branching is not performed according to the value of the packet counter Num as in the flowchart of FIG. 4, but is performed according to the value of the packet loss counter Loss. There is a possibility that it is more efficient to configure a flowchart that performs the processing. As an example, a packet loss counter Los
When the value of s becomes 5, the packet loss rate at that time may be calculated, and the packet loss rate may be used as the reception state information SI.

In the first and second embodiments, the IP network 12 compatible with the IP protocol is used. However, the network used in the present invention is not limited to the IP network.

The functions of the flowcharts in FIGS. 4, 5, and 7 may be described using any programming language, and may be realized in hardware using a logic circuit. .

[0158]

As described above, according to the present invention, a decrease in sound quality can be suppressed to a minimum level corresponding to network traffic, and high-quality and highly reliable voice communication can be realized. .

[Brief description of the drawings]

FIG. 1 is a schematic diagram illustrating an overall configuration of a communication system according to a first embodiment.

FIG. 2 is a schematic diagram showing a relationship between an average bit rate and sound quality.

FIG. 3 is a block diagram illustrating a schematic configuration of a main part of the first embodiment.

FIG. 4 is an operation explanatory diagram of the first embodiment.

FIG. 5 is an operation explanatory diagram of the first embodiment.

FIG. 6 is a block diagram illustrating a schematic configuration of a main part of the second embodiment.

FIG. 7 is an operation explanatory diagram of the second embodiment.

FIG. 8 is a schematic diagram illustrating an overall configuration of a communication system according to another embodiment.

FIG. 9 is a schematic diagram illustrating an overall configuration of a communication system according to another embodiment.

FIG. 10 is a schematic diagram illustrating an overall configuration of a communication system according to another embodiment.

[Explanation of symbols]

10, 40: communication system, 11: voice transmitting terminal, 12
... Network, 13 ... Speech receiving terminal, 20 ... Speech input unit, 21, 41 ... Speech coding unit, 21A, 43 ... Speech determination unit, 21B, 41B ... Speech frame coding unit, 21
C, 41C: silent frame encoding unit, 22: voice packet transmitting unit, 23: reception state information receiving unit, 24, 44: voiced judgment level calculation unit, 30: voice packet receiving unit, 31
... Voice decoding unit, 32 voice output unit, 33 reception state information creation unit, 34 reception state information transmission unit, 42 voice analysis unit, 45 voice determination Scale storage unit.

Claims (3)

[Claims] An audio packet receiving apparatus for receiving, via a predetermined network, an audio packet transmitted from an audio packet transmitting apparatus, detects a reception state of the audio packet that fluctuates according to traffic on the network. Receiving status detecting means, and receiving status information transmitting means for transmitting the receiving status information according to the receiving status detected by the receiving status detecting means to the network in order to deliver the receiving status information to the voice packet transmitting apparatus. Characteristic voice packet receiving device. 2. A voice packet transmitting apparatus for transmitting a voice packet to a voice packet receiving apparatus via a predetermined network, wherein a voice frame to be transmitted is voiced or silent in accordance with a determination level. Determination determining means; voice coding means for coding the silent frame determined to be silent by the determining means so as to have a smaller amount of information than the voiced frame determined to be voiced; Receiving status information receiving means for receiving, via the network, receiving status information transmitted by the voice packet receiving device according to the receiving status of the voice packet which varies correspondingly; A decision level changing means for changing the decision level in accordance with the received reception status information. The transmission device. 3. A packet communication system comprising: the voice packet receiving device according to claim 1; and the voice packet transmitting device according to claim 2.