JPH02205141A - Packet scrapping compensating method - Google Patents

Abstract

PURPOSE:To suppress deterioration in quality in the scrapping of a packet by classifying an Embeded coding system to an odd sampling system and an even sampling system, applying an interpolation processing on the odd or even sampling system transmitted by another packet when either packet is scrapped, and reproducing an Embeded code. CONSTITUTION:Inputted odd sampling system and even sampling system are packetized separately in sequence of odd and even numbers by a packetization part 130. The packet is sent on to a network, and is inputted to a packet disassembling part 140 at a reception side. A packet scrapping circuit 150 detects the presence/absence of the scrapping of the packet and which packet, the one of the odd sampling system or the one of the even sampling system, a scrapped packet is, and sends a result to a code system reproduction circuit 170. The code system reproduction circuit 170 reproduces a coding system of Embeded ADPCM.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a method for compensating for packet discard and cell discard in a voice packet communication system or an ATM communication system.

(Prior Art) A packet communication system is being realized in which an audio signal is encoded, then packetized, and communicated in units of packets.

Packet communication systems have the advantage of not only being able to centrally handle various media signals such as voice, images, and data, but also making efficient use of lines by utilizing the burst nature of voice signals to transmit only voiced sections. has. For this reason, packet communication and ATM communication are handled by l5DN, BIS
DN (7) It is attracting attention as a leader and research and development is being actively conducted.

However, in packet communication, packets are discarded when the network is congested or when packet delay is large, resulting in deterioration of voice quality. In particular, when ADPCM using adaptive prediction is used as the encoding method, quality deterioration is large when packets are discarded. Therefore, E
embeddedDPCM method is r Embedded DPC
M for varable bitrate tr
ansmlsslon J, (IEEE Trans
, C0M-28,7.

pp, 104G-1046Uly 1980))
(referred to as Document 1). Also, CCIT
T is CCITT 5GXVIII r Annex
to Question X/XV (Speech
Packet-1zat1on) Algorlt
hm and Protocol for 5peec
hPacketlzation J, (TD131.
Geneva 8-17 June 198g) (referred to as Document 2), tentatively recommends Embedded ADPCM as a coding method for voice packet communication as G, EMB, and recommends that the voice packet protocol be G, PVN.
This is a provisional recommendation as P.

4 and 5 are block diagrams of the encoder section and decoder section of the provisional recommendation G and EMB system. In the tJJ diagram, the human power of the encoder is μmPCM or A-PCM
This is an audio signal digitized by a codec.

610 converts the code of μmPCM or A-PCM into linear P
This is a PCN format converter that converts to CM code.

Further, 630 is an adaptive quantizer, and 670 is an adaptive predictor. The subtraction circuit 620 combines the human signal and the adaptive predictor 6
The difference between the predicted signal and the output of 70 is calculated and sent to the adaptive quantizer 630. The adaptive quantizer 630 quantizes the input prediction difference signal and outputs it as an ADPCM code. 64G is a bit mask circuit that masks the lower bits of the ADPCM output code by the maximum number of discardable bits and shifts it to the right. The output of the bit mask circuit 640 is sent as core bits to an adaptive dequantizer 650, which performs dequantization of the core pits. The output of the adaptive inverse quantizer is sent to the adaptive predictor 670 and the summing circuit 66.
Sent to 0. Addition circuit 660 also creates a locally decoded signal by adding the output signal of the adaptive inverse quantizer and the output signal of the adaptive predictor. The adaptive predictor 670 is an adaptive filter having a second-order pole and a sixth-order zero point, and generates a predicted signal by inputting the locally decoded signal and the dequantized predicted difference signal.

The number of bits in the adaptive quantizer 630 and the number of core bits fed back depend on the algorithm used. For example 3
2 Kbps (4,2) algorithm has a quantization of 4
The core pit is 2 bits in Bift. In FIG. 4, an adaptive quantizer 630 forms a feedforward path, and a bit mask circuit 640 and an adaptive inverse quantizer 650
Adaptive predictor 670 forms a feedback path.

Next, the operation of the decoder will be explained. The decoder in FIG. 5 has a feedback path including a bit mask circuit 680, a feedback adaptive inverse quantizer 690, and an adaptive predictor 710, and a feed forward adaptive inverse quantizer 72G and a PCM format conversion circuit 740, similar to Korg. It consists of a forward pass. The feedback path is exactly the same in the Korg and decoder.

The bit mask circuit 680 leaves the upper core pits of the manually generated ADPCM code, masks the lower bits, and shifts it to the right, thereby sending only the core pits to the feedback adaptive inverse quantizer 690. Feedback adaptive quantizer 690 performs inverse quantization of the core pits. Adaptive predictor 71
0 inputs the inverse quantized prediction/difference signal which is the output of 690 and the locally decoded signal which is the output of adder circuit 700, and outputs a prediction signal. Bit discard on the network is performed from the lower bits of the ADPCM code, and transmission of core pits is guaranteed. Therefore, the output of the bit mask circuit 680 on the decoder side is the same as the output of the bit mask circuit 640 on the Korg side.

Therefore, the outputs of the adaptive inverse quantizers 690, 650 and the adaptive predictors 710, 670 are exactly the same in the Coorg and decoder.

The feedforward adaptive quantizer 720 is an ADPCM
The core bits of the output code and the bits remaining without being discarded are dequantized. Addition circuit 730 adds the output of feedforward adaptive quantizer 720 and the output of adaptive predictor 710 to create a decoded signal. The obtained decoded signal is P
It is output to the CM format conversion circuit 740, where
A linear PCM code is converted to a μmPCM or A-PCM code. 750 is ADPCM-PCM-ADPC
This is a tandem connection correction circuit for preventing errors caused by synchronous tandem connections like M.

If bits of the output code are discarded in normal non-embedded ADPCM, the dequantized prediction difference signal will have different values between the CORG and the decoder.

As a result, the adaptive processing of the quantizer and the predictor becomes different asynchronous operations between the cog and the decoder, and errors due to discard are filtered by the synthesis filter, which increases quality deterioration due to bit discard.

On the other hand, in the Esbeded ADPCM described above, only the core pits are fed back to the predictor, so even if the lower bits other than the core pits are discarded on the network, asynchronous operation between the cog and the decoder does not occur.

Also, since the predicted signal is the same in the Korg and the decoder,
A quantization error corresponding to the number of discarded bits is simply added directly to the decoded signal, so there is little quality deterioration due to bit discard.

Document 2 describes a voice packet configuration method and protocol that take advantage of such characteristics of Eg+bedded ADPCM.

FIG. 6 shows the packet format described in Document 2. In the diagram, bit 1 is LSB, bit 8 is MSB
represents. P D (Protocol Discrlm
l-nator) is for distinguishing audio packets from other packets. BD I (Blo
ckDropping (ndlcator) indicates the number of blocks that can be discarded in the initial state of packetization and the number of blocks that can be discarded on each node of the network. Here, the block consists of 16 m5 (128
sample), the encoded output of the audio is expressed in bits as 1
This is information in units of 128 bits collected for frames. T
S (Time Stamp) indicates the cumulative sum of delay amounts occurring at each node of the network. CT (Coding)
Type ) is a field indicating the audio encoding method used when creating the packet.

S E Q (Sequence Number) is a number indicating the sequential order of packets, and is used when a packet is lost. N S (Notse Field)
is the background noise level t 7 e /l/d. N0N-DI? 0PPAB1 and EOCTETS are core pit blocks of Embedded ADPCM output, and contain information that cannot be discarded on the network. 0PTIONAL DROP PABLEB
LOCKS is a flag of the lower bits of the Embedded ADPCM, and is an information field that can be discarded if requested by the system on the network.

A layer 2 header and trailer are attached to the beginning and end of the packet. In a packet network protocol that uses packets with the format shown in Figure 6, packet discard is performed using 0PTIONALDROPPABL in the packet.
This is done by discarding the E BLOCK.

The above is the packet discard compensation method using the conventional Embedded ADPCM and packet format. This method causes less quality deterioration as described above when information is discarded within a packet, that is, in bit units. However, if discard occurs in packet units, E
Since the core bits of the mbed ADPCM are discarded, the quality deteriorates. Due to packet discard, one frame (16as) of signal is completely lost, making it impossible to reproduce the original audio signal. This state does not end in one frame, and due to the asynchronous operation of the encoder and decoder,
It will last for more than one frame. As a compensation method for discarding packets, there is a method of interpolating and reproducing the signals of packets before and after the discarded packet, but since the predicted difference signal output from ADPCM is a signal from which correlation has been removed, one frame ( Even if interpolation is performed using separated samples (128 samples), there is almost no interpolation effect and quality deterioration is inevitable.

(Problem to be Solved by the Invention) In the conventional packet discard compensation method using Embedded ADPCM, when packet discard occurs, the Embedded
Since the core pit of eded ADPCM is also discarded,
The problem is that the original audio signal cannot be reproduced and the encoder and decoder operate asynchronously, resulting in significant quality deterioration.

The present invention has been made in view of these problems, and an object of the present invention is to provide a packet discard compensation method that causes less quality deterioration even when packets are discarded.

[Structure of the Invention] (Means for Solving the Problems) The present invention provides means for dividing an output sample sequence of an Embedded ADPCM into an odd sample sequence and an even sample sequence, and a means for separately packetizing the odd and even sample sequences. and a means to know which packets have been discarded on the receiving side, and if there is no discarded packet, E.
means for reproducing an output sample sequence of the mbed ADPCM; and when one of the odd sample sequence packet and the even sample sequence packet is discarded, interpolation processing is performed on the odd or even sample sequence of the other packet; Embedded ADPCM has a means for reproducing an output sample sequence of Embedded ADPCM.

(Function) Embedded ADPCM is an encoding method that causes less quality deterioration when discarding bit units, that is, discarding blocks within a packet. In the present invention, Embedded ADPCM is used as a means for encoding audio. Embedded
The means for dividing the ADPCM output sample sequence into an odd sample sequence and an even sample sequence is a means used as a preprocessing means for separately packetizing the odd sample sequence and the even sample sequence. Transmitting the odd-numbered sample series and the even-numbered sample series as separate packets causes Esbeded ADPC in a certain section by discarding the packets.
This has the effect of greatly reducing the probability that the code of M will disappear completely. On the receiving side, there is a means to know about discarded packets and an Emb.
There is an eded ADPCM code sequence reproducing means, and the output of the means for knowing discarded packets is esbaded ADPCM.
The code sequence is sent to the code sequence reproducing means.

The E+gbeded ADPCM code sequence reproducing means directly converts the Embedded code from odd sample sequence packets and even sample sequence packets when there is no packet discard.
The ADPCM code sequence is reproduced, and when one of the odd sample sequence packet and the even sample sequence packet is discarded, interpolation processing is performed on the sample sequence of the other packet to reproduce the Esbeded ADPCM code sequence. This E-beded ADPCM code sequence regeneration method is different from the conventional Embedded ADPCM code sequence regeneration method when there is no discard in units of packets and only discards in units of blocks within packets.
d Has the same excellent characteristics as ADPCM, with little quality deterioration. Even if packets are discarded, the prediction is incomplete, so there is a correlation between adjacent samples in the output sample sequence of Embedded ADPCM, and it is difficult to change the output sample sequence from an odd sample sequence to an even sample sequence or from an even sample sequence to an odd sample sequence. Interpolation playback is possible, with little quality deterioration.

(Example) Hereinafter, an example according to the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of a voice packet communication system according to an embodiment of the present invention.

In FIG. 1, 100 is an input terminal, and μ-PCM
Alternatively, the active portion of the audio signal digitized by the A-PCM codec is input.

110 is Embedded ADPCM -r-n:7-
The conventional Embedder, whose block diagram is shown in Fig.
ded ADPCM encoder. 120 is Emb
A code sequence regeneration circuit that decomposes an eded ADPCM code sequence into an odd sample sequence and an even sample sequence,
Its output is sent to the packetizer. The packetization unit separately packetizes the input odd sample series and even sample series in the order of odd numbers and even numbers in units of 128 samples in the format shown in FIG. The packet is sent out on the network and manually input to the packet disassembly unit 140 on the receiving side. The packet decomposition unit 140 decomposes the packet into each piece of information. 150 is a packet discard detection circuit, and based on the SEQ information in the format shown in Fig. 6,
The presence or absence of packet discard is detected, and whether the discarded packet is an odd sample series packet or an even sample series packet, and the result is sent to the code series regeneration circuit 170.
send to The buffer circuit 160 includes the bucket decomposition unit 140
This is a memory circuit that stores at least three packets of the Embedded ADPCM code sequence obtained in the above. The code sequence reproduction circuit 170 is a circuit that reproduces the code sequence of EIlbedADPCM, and performs the processing shown in the flowchart of FIG. When the output of the packet discard detection circuit indicates that there is no packet discard, the E
Regenerate the mbed ADPCM code sequence. If a packet is discarded and the discarded packet is an odd sample series packet, an Ea+bcded ADPCM code series is reproduced by performing interpolation processing on the even sample series.

If the discarded packet is an even sample series packet, Em
The code sequence of the bedded ADPCM is reproduced.

FIG. 3 is a block diagram showing an example of the configuration of the interpolation processing circuit. In this circuit, the remaining odd or even sample series that have not been discarded are input from the input terminal 200, and the switch circuit 210 inserts 0 between the samples. switch circuit 21
O is input to the terminal 220 of 0. The signal sequence with zeros inserted between samples is input to a low-pass filter with a bandwidth of 2 KHz. The odd sample series and even sample series of Embedded ADPCM are Etibe
ded ADPCM code sequence spectrum 2KII
It has a spectrum folded at z. The 211 Kz low-pass filter is used to reproduce the spectral components of the Embedded ADPCM code sequence up to at least 211 Kz. As a method of interpolation processing, average value interpolation may be used in addition to a method using a low-pass filter.

That is, the odd or even sample sequence is d (n-1)
, d Sample d discarded as (n+1)
(n) is interpolated using the following equation.

d (n) = ”-t d (n-1) + d (
nil) 1 Also, find the optimal interpolation coefficient on the transmitting side,
The coefficients may be used for interpolation.

The output of the code sequence regeneration circuit is the Embedded ADPC
The audio signal is sent to M decoder 180, where the audio signal is decoded. The EIlbed ADPCM decoder is a conventional Embedded ADPCM decoder having the configuration shown in FIG.

In this way, in the packet communication system of this embodiment, even if a packet is discarded, the EmbeddedADPC
Since M code sequences can be reproduced, there is little quality deterioration. Since interpolation reproduction is performed by a low-pass filter, the quality of the decoded audio is only slightly suppressed, with high frequency components being slightly suppressed.

In the above embodiment, the Eibed ADPCM code sequence is divided into two, an odd sample sequence and an even sample sequence, but it may be divided into any number of sequences of 2 or more.

[Effects of the Invention] As stated above, in the present invention, the EmbeddedAD
Since audio encoding is performed using PCM, there is not only little quality deterioration due to information discarded in bit units, that is, information discarded in packets, but also Embedded ADP
The CM code sequence is divided into an odd sample sequence and an even sample sequence, each of which is sent in a separate packet, and if either the odd or even sample sequence packet is discarded, interpolation is performed from the sample sequence of the remaining packet. Since the sample sequence discarded by processing is regenerated, there is an effect that there is little quality deterioration even when packets are discarded.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a voice packet communication system according to an embodiment of the present invention, FIG. 2 is a flowchart showing the operation of the code sequence reproducing circuit shown in FIG. 1, and FIG. 3 is an embodiment of the present invention. A block diagram showing a configuration example of an interpolation processing circuit according to an example, FIG. 4 is a block diagram of an encoder section of a conventional Eibed ADPCM, and FIG. 5 is a block diagram of a conventional Embedded ADPCM.
d The block diagram of the ADPCM decoder section, Figure 6, is as follows:
FIG. 2 is a diagram showing a conventional packet format. 100.200.600...Input terminal, 110-Em
beded ADPCM encoder section code 20... Code sequence decomposition circuit, 130... Packetization section, 140... Packet disassembly section, 150... Packet discard detection circuit, 160... Buffer circuit, 170... Code sequence reproduction circuit, 180-Embedded ADPCM decoder section, 21
0...Switch circuit, 230...Low pass filter 610...PCM format conversion circuit, 620...
- Subtraction circuit, 630...Adaptive quantizer, 640.680...Bit mask circuit, 650...
Adaptive inverse quantizer, 660.700,730...addition circuit, 670.71
0...Adaptive predictor, 690...Feedback adaptive inverse quantizer, 720...
...Feedforward adaptive inverse quantizer, 740...
PCM format conversion circuit, 750... tandem connection correction circuit.

Claims (2)

[Claims] (1) In an audio packet communication system in which the active portion of an audio signal is also transmitted in packet units, an Embedded
It has means for performing ed encoding, means for dividing the embedded code sequence into an odd sample sequence and an even sample sequence, and means for separately packetizing and transmitting the odd and even sample sequences, and the receiving side decomposes the packets. 1. A packet discard compensation method comprising: a means for determining discarded packets; a means for accumulating received packets; a means for reproducing an embedded code sequence; and a means for performing embedded decoding. (2) When there is no packet discard, the means for reproducing the embedded code sequence is to reproduce the embedded code sequence using the packets of the odd sample sequence and the packets of the even sample sequence, and to reproduce the embedded code sequence using the packets of the odd sample sequence and the packets of the even sample sequence. The packet according to claim 1, characterized in that the packet has a function of performing interpolation processing on the odd or even sample sequence transmitted in the other packet to reproduce the embedded code sequence when one of the packets is discarded. Disposal compensation method.