JP2005107283A - Method, device and program of packet loss concealment in voip voice communication - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device of packet loss concealment in VoIP communication which can effectively perform error concealment by waveform duplication of a packet lost in the VoIP voice communication with high accuracy in order to suppress deterioration of call quality in VoIP as much as possible. <P>SOLUTION: When the device detects the packet lost in reception of packets transmitted via an IP network, calculates pitches of frames before and after a lost part in the lost packet (steps S17, S19), calculates pitch fluctuation rate from the pitches of the frames before and after the lost part (a step S21), compares the pitch fluctuation rate with a predetermined threshold (a step S23), when the pitch fluctuation rate is larger than the predetermined threshold as a result of this comparison, executes a regular 2-side PWR(Pitch Waveform Replication) method (a step S25) and when the pitch fluctuation rate is smaller than the predetermined threshold, executes a 2-side PWR method in consideration of the pitch fluctuation rate (a step S27). <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a packet loss concealment method, apparatus, and program in VoIP voice communication that conceals a packet lost in VoIP (Voice over Internet Protocol) voice communication using an IP network by waveform duplication.

  In recent years, VoIP technology that performs voice communication using an IP network has attracted attention (see, for example, Non-Patent Document 1). The VoIP telephone service has been rapidly spreading in recent years because it can realize an inexpensive communication charge compared to the conventional telephone service that sets a call charge proportional to distance and time. (For example, Non-Patent Document 2).

  However, while the conventional telephone service realizes voice communication using a guarantee type network, VoIP realizes voice communication using a best effort type IP network that is inherently unsuitable for real-time communication. Communication errors such as disappearance and delay are unavoidable, and there is a problem in principle that the call quality is likely to deteriorate (for example, Non-Patent Documents 3 and 4).

  In realizing voice communication by VoIP, it is important to prepare a network so that packets are received reliably within a certain period of time in order to prevent the above-described errors as much as possible (for example, Non-patent document 5). However, it is impossible to completely control a best effort network, and it is difficult to eliminate all errors. For this reason, it is necessary to take measures for minimizing the deterioration of the call quality as much as possible even in such a case, assuming that a certain degree of error occurs. Normally, in non-real-time communication, when an error occurs, the error is dealt with by retransmitting the packet. However, in VoIP that requires real-time performance, there is almost no time to retransmit the packet. It is necessary to perform error concealment processing that is not necessary.

  Various methods have been proposed for error concealment processing in the past, but can be broadly classified into two types: a sender-based method for dealing with the transmitting side and a receiver-based method for dealing with the receiving side. (Non-Patent Documents 6 and 7). There are three types of error situations: instantaneous interruption due to loss of a single packet, overrun due to reception buffer overflow, and underrun due to reception buffer underflow.

  In VoIP voice communication, analog voice is AD-converted on the transmission side during voice call, and digitized voice data is divided into frames, and then compression processing is performed. Then, an IP packet is created by mounting the compressed audio data in the payload and transmitted to the receiving side via the IP network. When receiving the IP packet, the receiving side reproduces the analog voice by decomposing the IP packet, decompressing the compressed voice data, and DA conversion.

  In VoIP, UDP (User Datagram Protocol) is used as a transport layer protocol to realize real-time communication. Therefore, even if a communication error such as packet loss or delay occurs, the packet is not retransmitted. In general, the frame length of the compressed audio data mounted on the payload is generally set to 20 ms. In addition, although the types of codecs supported differ depending on the system, the codecs have become G. 711 is generally used. G. Reference numeral 711 denotes a codec that logarithmically quantizes voice data digitized at a sampling frequency of 8 kHz to a quantization accuracy of 8 bits. Although the compression efficiency is lower than that of other codecs, it is possible to suppress deterioration in speech quality. . In addition, G. There are two types of 711, μ-1aw and A-1aw.

  In the waveform replication method, which is a conventional receiver-based error concealment process, error concealment is performed by estimating a replacement block serving as a substitute for a lost frame from normally received speech data and copying this to the lost frame. The waveform replication method is classified into a WR (Wave Replication) method and a PWR (Pitch Waveform Replication) method depending on the definition of the replacement block.

  In the WR method, as shown in FIG. 9 (a), a portion having the highest correlation is identified by template matching, and the voice data immediately after this is used as a replacement block, and this replacement block is copied as a duplicate block to the lost frame at once. Error concealment. In the PWR method, as shown in FIG. 9B, a one-pitch waveform is identified by template matching, this is used as a replacement block, and this replacement block is periodically copied as a duplicate block to a lost frame. Perform error concealment.

  As shown in FIG. 9, the frame length is L, the search window length is M, the template length is N, and the received voice data is s (n). At this time, the kth (≧ 0) th packet is momentarily interrupted. Assuming that the template matching in the WR method is performed, the time m giving the maximum value of the cross-correlation function C (m) defined by the equation (1) is obtained.

  On the other hand, in the template matching in the PWR method, the time m giving the maximum value of the cross-correlation function C (m) defined by the equation (2) is obtained.

  In the WR method, the length of audio data necessary for processing is L + M + N, and at least two packets immediately before the lost frame need to be normally received. On the other hand, in the PWR method, the length of audio data necessary for processing is M + N, and the maximum pitch length that can be extracted is equal to the search window length M. In general, the pitch length of male voices is distributed from 5 ms to 12 ms, and the pitch length of female voices is distributed from 2 ms to 7 ms. Therefore, if the frame length is L = 20 ms and the template length is N ≦ 8 ms, the pitch length is just before the loss frame. A pitch waveform up to a pitch length of 12 ms can be extracted with only one received packet.

  When the frame length is short, error concealment can be effectively performed even with the WR method. However, with the WR method, if the audio data required for processing becomes long, the steadiness of the audio data cannot be sufficiently guaranteed, so the frame When the length is long, it is more effective to apply the PWR method.

  Basically, in the WR method and the PWR method, error concealment is performed using voice data immediately before the lost frame. However, since VoIP absorbs delay jitter, a reception buffer of about 3 packets is set. The error concealment with higher accuracy can be realized by extending the processing to 2-side processing using the audio data before and after. In particular, in the PWR method, the length of voice data required for processing can be suppressed to one packet, so that the look-ahead delay can be suppressed to one packet even when the processing is expanded to 2-side processing. High practicality.

  In the 2-side PWR method, as shown in FIG. 10, the PWR method is executed by using one packet before and after the lost frame, and error concealment is performed by overlapping both of them, thereby starting and ending the speech waveform. It is possible to improve the accuracy of error concealment in a portion where the non-stationary property such as the portion is remarkable (Non-patent Document 8).

  The conventional 2-side PWR method will be described in detail with reference to FIG. As a result of receiving the original voice as shown in FIG. 10A on the receiving side, if the k-th frame cannot be received and is lost as shown in FIG. As shown in FIG. 10C, a backward PWR is performed in which a replacement block is estimated from the front k-1 number frame and copied to the lost frame. Further, as shown in FIG. A forward PWR is performed in which a replacement block is estimated from the frame and copied to the lost frame.

  Then, as shown in FIG. 10 (e), the replacement block from the front side of FIG. 10 (c) and the replacement block from the rear side of FIG. 10 (d) are overlapped and added by proportional distribution. Then, error concealment speech is generated. As a result, the lost frame is reproduced between the front frame and the rear frame, and is reproduced as a continuous sound with no loss.

The reproduced error concealed speech shown in FIG. 10 (e) is considerably different from the original speech so that a differential waveform as shown in FIG. 10 (f) is generated when compared with the original speech shown in FIG. 10 (a). The waveform of the replacement block generated from the previous frame is out of phase with the waveform of the replacement block generated from the rear frame, and the pitch of both waveforms is shifted. Because.
Hiroshi Fujiwara, multimedia information compression, Kyoritsu Publishing, 2000 G, Held, Voice & Data Integration Technology Guide, Impress, 2000 Keiichi Imai, “Problems in Realizing VoIP”, IEICE Journal, vol.83, no.4, pp.295-301, 2001 Hiromi Nagahama, “Problems on VoIP Quality”, IEICE Technical Report, vol.IN2000-128,2000 Satoshi Toda, Detailed Network QoS Technology, Ohmsha, 2001 C. Perkins, O. Hodson and V. Hardman, “A survey of packet loss recovery techniques for streaming audio”, IEEE Network Magazine, pp. 40-48, September / October 1998 H. Sanneck, “Packet Loss Recovery and Control for Voice Transmission over the Internet”, Ph.D. thesis, Technical University Berlin, 2000 Nobuko Komaki, Naofumi Aoki, Tsuyoshi Yamamoto, “A Method for Concealing Packet Loss in VoIP Based on Waveform Replacement”, IEICE Tech. Bulletin, vol.CQ2002-59,2002

  As described above, since VoIP realizes voice communication using a best-effort IP network that is inherently unsuitable for real-time communication, communication errors such as packet loss and delay are unavoidable, leading to a decrease in call quality. The 2-side PWR method, which is a conventional error concealment method using waveform duplication that conceals such packet loss, causes a phase shift in the waveform generated from the previous and subsequent frames. The error concealed speech is considerably different from the original speech such that a relatively large difference waveform as shown in FIG. 10 (f) is generated with respect to the original speech, and the error concealment cannot be performed effectively. There is.

  The present invention has been made in view of the above, and an object thereof is to effectively conceal an error by duplicating a waveform of a packet lost in VoIP voice communication with high accuracy in order to suppress a decrease in call quality in VoIP as much as possible. It is an object to provide a packet loss concealment method, apparatus, and program for VoIP voice communication that can be performed in the following manner.

  A packet loss concealment method in VoIP voice communication according to claim 1 of the present invention is a packet loss concealment method in VoIP voice communication that conceals a lost packet in VoIP voice communication using an IP network by waveform duplication. Detects a lost packet in the reception of a packet transmitted through the network, calculates the frame pitch before and after the lost part in the detected lost packet, and based on the calculated previous and next frame pitch The pitch fluctuation rate between frames is calculated, and the calculated pitch fluctuation rate is compared with a predetermined threshold. If the pitch fluctuation rate is larger than the predetermined threshold as a result of this comparison, a normal 2-side RWR (Pitch Waveform Replication ) Method, and when the pitch variation rate is smaller than a predetermined threshold, the pitch The gist is to implement the 2-side PWR method considering the fluctuation rate.

  The packet loss concealment device in VoIP voice communication of the present invention according to claim 2 is a packet loss concealment device in VoIP voice communication that conceals a lost packet in VoIP voice communication using an IP network by waveform duplication. Lost packet detection means for detecting a lost packet in reception of a packet transmitted via the IP network, pitch calculation means for calculating the pitch of frames before and after the lost portion in the detected lost packet, and this calculation A pitch fluctuation rate calculating means for calculating a pitch fluctuation rate between the preceding and succeeding frames based on the pitches of the preceding and following frames, a comparing means for comparing the calculated pitch fluctuation rate with a predetermined threshold, and as a result of this comparison, the pitch When the rate of change is greater than a predetermined threshold, normal 2-side PWR Pitch Waveform Replication) technique carried out, when the pitch variation rate is smaller than a predetermined threshold value, and summarized in that to implement the 2-sidePWR method considering the pitch variation rate.

  A packet loss concealment program in VoIP voice communication according to claim 3 of the present invention is a packet loss concealment in VoIP voice communication that can be executed by a computer for concealing a lost packet in VoIP voice communication using an IP network by waveform duplication. A program for calculating lost packet detection means for detecting a lost packet in receiving a packet transmitted from the computer via the IP network, and calculating a pitch of frames before and after the lost portion in the detected lost packet A pitch calculating means for calculating the pitch fluctuation rate between the preceding and following frames based on the calculated pitch of the preceding and following frames, and a comparing means for comparing the calculated pitch fluctuation rate with a predetermined threshold value As a result of this comparison, When the fluctuation rate is larger than a predetermined threshold, a normal 2-side PWR (Pitch Waveform Replication) method is performed. When the pitch fluctuation rate is smaller than the predetermined threshold, the 2-side PWR method considering the pitch fluctuation rate is performed. The gist is to function as PWR implementation means.

  According to the present invention, the pitch of the frames before and after the lost part in the lost packet detected in the reception of the packet is calculated, the pitch variation rate is calculated based on the pitch of the preceding and following frames, and this pitch variation rate is determined in advance. When the pitch fluctuation rate is larger than the predetermined threshold value, the normal 2-side PWR method is performed. When the pitch fluctuation rate is smaller than the predetermined threshold value, the pitch fluctuation rate is considered. Since the 2-side PWR method is implemented, phase shift can be suppressed and error concealment can be effectively performed with high accuracy.

   The packet loss concealment method in the VoIP voice communication according to the present invention calculates the pitch of frames before and after the lost portion in the detected lost packet when detecting the lost packet in receiving the packet transmitted through the IP network. Then, the pitch fluctuation rate is calculated based on the pitches of the preceding and following frames, the pitch fluctuation rate is compared with a predetermined threshold value, and if the pitch fluctuation rate is larger than the predetermined threshold value as a result of this comparison, the normal 2- When the side PWR method is performed and the pitch variation rate is smaller than a predetermined threshold, the 2-side PWR method is performed in consideration of the pitch variation rate.

FIG. 1 is a flowchart showing a processing procedure of a packet loss concealment method in VoIP voice communication according to an embodiment of the present invention.
The packet loss concealment method in the VoIP voice communication of the present embodiment uses the conventional 2-side PWR method described in FIG. 10, but in this 2-side PWR method, the phase shift is considered in consideration of the pitch variation before and after the lost frame. Thus, the difference between the copied error concealed speech and the original speech is reduced, so that concealment by waveform duplication of the lost packet can be effectively performed with high accuracy.

  More specifically, the packet loss concealment method in VoIP voice communication according to the present embodiment is effective in error concealment processing by waveform duplication in order to suppress the deterioration of communication quality due to communication errors such as packet loss and delay in VoIP voice communication as much as possible. However, in this embodiment, a receiver-based method to be dealt with on the receiving side is used as an error concealment process, and the receiver-based method is highly accurate despite relatively simple processing. In order to perform error concealment more effectively with this method of error concealment using waveform duplication, which is based on the waveform duplication method capable of error concealment, in 2-side processing using audio data before and after the lost frame Using the extended 2-side PWR method and taking into account fluctuations in pitch extracted when performing error concealment from the 2-side, phase shift To reduce and performs error concealment processing from the front and rear with high accuracy.

With reference to FIG. 1, a packet loss concealment method in VoIP voice communication according to an embodiment of the present invention will be described in detail.
In this embodiment, although the 2-side PWR method is used, in order to further improve the accuracy of the 2-side PWR method, interpolation is performed in consideration of the pitch variation of the audio data before and after the lost frame. In this embodiment, the template length in the 2-side PWR method is set to N = 8 ms, the search window length is set to M = 10 ms so that the minimum value of the pitch length is 2 ms and the maximum value is 12 ms.

  In FIG. 1, the receiving side that receives a packet transmitted via the IP network in VoIP voice communication acquires the k + 1th packet (step S11), and acquires the voice data of the k + 1th frame (step S13). ) Here, a check is made to determine whether or not the k-th packet has been lost, that is, whether the k-th packet has been lost by the lost packet detection means and has not been received normally (detection of a lost packet). (Step S15).

  When the k-th packet is lost and is not normally received, the pitch calculation means first extracts the pitch Pb of the k-1th frame, which is the previous frame (step S17), and one more The pitch Pf of the (k + 1) th frame that is the subsequent frame is extracted (step S19). Next, the pitch fluctuation rate calculating means calculates the pitch fluctuation rate α from the extracted pitches of the (k−1) th and k + 1th frames according to the following equation (step S21).

α = (P _f −P _b ) / L (3)
Here, L is the frame length.
The pitch variation rate α will be described with reference to FIG. In FIG. 2, it is assumed that the frame length is L and the kth frame is a lost frame, and the pitch variation rate of the lost frame is calculated. The pitch of the (k-1) th frame is Pb And the pitch of the (k + 1) th frame is Pf , The pitch variation rate α in the k-th frame between the two frames is calculated as shown in the equation (3), which is calculated by using both pitches Pb. , Pf This represents the slope of the line segment defined by. Therefore, the pitch length at each time point of the kth frame can be calculated from the pitch variation rate α. In other words, the pitch length of the kth frame is obtained by multiplying the pitch variation rate α by the time n from the start time n = 0 of the kth frame and the pitch length Pb at the start time n = 0 of the kth frame. It can be calculated by adding.

  In the steady section of the voiced sound, it is expected that the pitch lengths of the frames before and after the loss frame take similar values. In this case, the absolute value of the pitch variation rate α is a small value. However, in the case of transition from silent sound or unvoiced sound to voiced sound or vice versa, non-stationarity is strong, and the absolute value of the pitch fluctuation rate does not always take a small value.

  FIG. 3 is a histogram of absolute values of the pitch fluctuation rate α extracted from the voice data of 10 male and female voices. The vertical axis represents the number of packets, and the horizontal axis represents the absolute value of the pitch variation rate α. Note that the audio data is obtained by re-sampling audio material randomly selected from the audio database of the Acoustical Society of Japan with a sampling frequency of 8 kHz and a quantization accuracy of 16 bits. The total time length of the voice data was 89.92 s, and the number of packets was 4496. As shown in FIG. 3, it can be seen that the absolute value of the pitch fluctuation rate α in normal audio data may be concentrated near 0 and there is no significant fluctuation.

  Returning to the flowchart shown in FIG. 1, in this embodiment, the comparison means sets a threshold value T for the absolute value of the pitch fluctuation rate α, and whether or not the absolute value of the pitch fluctuation rate α is smaller than the threshold value T. Is determined (step S23). When the absolute value of the pitch fluctuation rate α is larger than the threshold value T, the conventional 2-side PWR method is performed, and the replacement block from the front side and the rear side are changed without changing the pitch length as described in FIG. Extrapolation is performed by overlapping and adding the replacement block from the side (step S25).

  In other cases, that is, when the absolute value of the pitch fluctuation rate α is smaller than the threshold T, the lost frame is regarded as a steady section of voiced sound, and the 2-side PWR method is performed in consideration of the pitch fluctuation, and the replacement block is replaced with the lost frame. Each time the pitch waveform is copied, the pitch length of the pitch waveform at the copy start time n (n = 0 at the start time of the lost frame) is calculated by equation (4) in the backward PWR of the replacement from the previous frame, and later In the forward PWR for replacement from the side frame, the pitch length of each pitch waveform is updated by calculation using equation (5) (step S27).

FIG. 4 explains the 2-side PWR method in consideration of the pitch variation rate α in step S27 with reference to the waveform diagram shown in FIG.
As a result of receiving the original speech as shown in FIG. 4A on the receiving side, if the k-th frame cannot be received and is lost as shown in FIG. As shown in FIG. 4C, a backward PWR is performed in which a replacement block from the frame of the k-1 number on the front side is estimated and copied to the lost frame. Further, as shown in FIG. The forward PWR in which the replacement block is estimated from the side and copied to the lost frame is performed. The backward PWR and the forward PWR are calculated by calculating the pitch length for each copy by the equations (4) and (5), respectively. It is performed so as to have a pitch length. As a result, as shown by the vertical lines across FIGS. 4C and 4D, copying is performed so that there is no phase shift between the two waveforms and the phases almost coincide.

  As shown in FIG. 4 (e), both the backward PWR and the forward PWR that are generated in such a manner that the pitch length is adjusted and the phases are substantially matched and copied to the lost frame take into account the time from the copy start time. Overlap and add by the proportional distribution, error concealment speech is generated. As a result, the lost frame is continuous between the front frame and the rear frame, and is reproduced as a lossless sound in consideration of pitch fluctuation.

  When the error concealed speech reproduced as shown in FIG. 4 (e) is compared with the original speech shown in FIG. 4 (a), a differential waveform as shown in FIG. 4 (f) is generated. This difference waveform is considerably reduced compared to the conventional difference waveform shown in FIG. 10E, and it can be seen that the error concealed speech is much closer to the original speech. This is because, as a result of performing the 2-side PWR in consideration of the pitch variation, the phase shift caused by the pitch variation in the steady section of the voiced sound that has been ignored in the conventional 2-side PWR method is reduced.

  As described above, when the pitch fluctuation rate α is larger than the threshold value, the comparison unit performs the conventional 2-side PWR method, and when it is smaller, the comparison unit performs the 2-side PWR method considering the pitch variation. Thus, a replacement waveform for the lost block is generated (step S29), the audio data of this replacement waveform is inserted between frames before and after k-1 and k + 1, and the frames are connected (step S31) to reproduce the audio data. (Step S33).

  5 and 6 are graphs showing the relationship between the threshold value T and the quality compared with the pitch variation rate α in step S23 of FIG. Both figures show threshold values of SNR (Signal-to-Noise Ratio) and PESQ (Perceptual Evaluation of Speech Quality) of error concealed speech when the packet loss concealment method of the present invention is applied to all frames in the speech data described above. It is a graph shown as quality with respect to T. As shown in both figures, it is understood that the threshold value T is relatively small, for example, about 0.4 to 0.15, and if it is too large, the quality of SNR, PESQ, etc. deteriorates.

  PESQ is an index that takes into account the pinion model, has a high correlation with subjective evaluation, and is applied to voice data quality evaluation in VoIP. The value of PESQ is distributed in the range from 4.5 to -0.5, and the larger the value, the better the quality. As a result, if the threshold value is increased too much, the quality may be lowered, and it is important to set the threshold value appropriately.

  FIGS. 7 and 8 show the SNR and PESQ with respect to changes in the packet loss rate as a result of evaluation experiments performed to confirm the effectiveness of the packet loss concealment method of the present invention described above in comparison with the conventional method. It is a graph. Note that the packet loss rate was varied from 0.5 to 10%. In addition, SNR and PESQ use the average values obtained from all audio data as final evaluation values.

  The audio data used in this evaluation experiment was resampled with audio data of 4 male speakers and 4 female speakers randomly selected from the audio database of the Acoustical Society of Japan at a sampling frequency of 8 kHz and a quantization accuracy of 16 bits. Is. However, in consideration of the minimum value of the packet loss rate that is a parameter of the evaluation experiment, 1 is obtained by randomly connecting audio materials for each speaker so that the time length is 20 s and the number of packets is 1000 or more. A total of 40 voice data was prepared for each of five speakers. As a result, the total time length of the voice data was 898.16 s, and the number of packets was 44908.

  In the evaluation experiment, a pseudo error is generated in the voice data, and (a) the packet loss concealment method of the present invention, (b) the conventional 2-side PWR method, (c) G.G. For each of the 711PWR methods, an objective quality evaluation was performed using SNR and PESQ as indices. Note that the threshold in the packet loss concealment method of the present invention is T = 0.1. In the evaluation experiment, only the instantaneous interruption due to the loss of a single packet generated randomly was targeted. Furthermore, even if the packet loss rate is the same, errors are generated in order from the frame with the largest gain for each audio data in consideration of the fact that the evaluation varies depending on the magnitude of the original gain of the lost frame.

  7 and 8, the SNR and PESQ according to the packet loss concealment method of the present invention shown in (a) are the same as the conventional 2-side PWR method shown in (b) and (c). It was found that the value was larger than that according to the 711 PLC (Packet Loss Concealment) method, and error concealment was more effective than other conventional methods.

  The processing procedure of the packet loss concealment method in the VoIP voice communication of the above embodiment is recorded as a program on a recording medium such as a CD or FD, and this recording medium is incorporated in a computer system or recorded on a recording medium. It is possible to function as a packet loss concealment device that implements a packet loss concealment method by downloading a program to a computer system via a communication line or installing it from a recording medium and operating the computer system with the program. Of course, the use of such a recording medium can improve the distribution.

It is a flowchart which shows the process sequence of the packet loss concealment method in the VoIP voice communication concerning one Example of this invention. It is a figure for demonstrating the update of the pitch length in the packet loss concealment method of the Example shown in FIG. It is a histogram which shows the number of packets with respect to the absolute value of pitch variation rate (alpha). It is a wave form diagram for demonstrating the 2-sidePWR method in consideration of the pitch fluctuation | variation in the packet loss concealment method of the Example shown in FIG. It is a figure which shows SNR with respect to the threshold value compared with a pitch fluctuation rate. It is a figure which shows PESQ with respect to the threshold value compared with a pitch fluctuation rate. It is a graph which shows SNR with respect to the change of the packet loss rate of the evaluation experiment result performed in order to confirm the effectiveness of the packet loss concealment method of this invention compared with the case of a conventional method. It is a graph which shows PESQ with respect to the change of the packet loss rate of the evaluation experiment result performed in order to confirm the effectiveness of the packet loss concealment method of this invention compared with the case of a conventional method. It is a figure for demonstrating the conventional WR method and PWR method. It is a figure for demonstrating the conventional 2-sidePWR method.

Claims (3)

A packet loss concealment method in VoIP voice communication for concealing a lost packet in VoIP voice communication using an IP network by waveform duplication,
Detect lost packets in receiving packets sent over the IP network,
Calculate the frame pitch before and after the lost part in the detected lost packet,
Based on the calculated pitch of the previous and next frames, the pitch fluctuation rate between the previous and next frames is calculated,
Compare this calculated pitch fluctuation rate with a predetermined threshold,
As a result of this comparison, when the pitch fluctuation rate is larger than a predetermined threshold, the normal 2-side RWR method is executed, and when the pitch fluctuation rate is lower than the predetermined threshold, the 2-side PWR method considering the pitch fluctuation rate is executed. A packet loss concealment method in VoIP voice communication. A packet loss concealment device in VoIP voice communication that conceals a lost packet in VoIP voice communication using an IP network by waveform duplication,
A lost packet detecting means for detecting a lost packet in receiving a packet transmitted via the IP network;
A pitch calculating means for calculating the pitch of the frames before and after the lost portion in the detected lost packet;
A pitch variation rate calculating means for calculating a pitch variation rate between the preceding and following frames based on the calculated preceding and following frame pitches;
A comparison means for comparing the calculated pitch fluctuation rate with a predetermined threshold;
As a result of this comparison, when the pitch fluctuation rate is larger than a predetermined threshold, the normal 2-side PWR method is executed, and when the pitch fluctuation rate is lower than the predetermined threshold, the 2-side PWR method considering the pitch fluctuation rate is executed. A packet loss concealment device for VoIP voice communication. A packet loss concealment program in VoIP voice communication that can be executed by a computer for concealing a lost packet in VoIP voice communication using an IP network by waveform duplication,
Lost packet detection means for detecting lost packets in receiving packets transmitted from the computer via the IP network;
A pitch calculating means for calculating the pitch of the frames before and after the lost portion in the detected lost packet;
A pitch variation rate calculating means for calculating a pitch variation rate between the preceding and following frames based on the calculated preceding and following frame pitches;
A comparison means for comparing the calculated pitch fluctuation rate with a predetermined threshold;
As a result of this comparison, when the pitch fluctuation rate is larger than a predetermined threshold, the normal 2-side PWR method is executed, and when the pitch fluctuation rate is lower than the predetermined threshold, the 2-side PWR method considering the pitch fluctuation rate is executed. A packet loss concealment program in VoIP voice communication, which functions as 2-side PWR implementation means.

Images