JPH0223744A - Sound packet interpolation system - Google Patents

Abstract

PURPOSE:To attain interpolation without damaging the praticality of a sound packet even if a packet lacks by using the weighing mean value of a prediction value and an interpolation value as the sample value of a connection part before and after the lack packet. CONSTITUTION:If a PCM signal x(m) being being the sound packet is inputted from an input terminal 101 to a waveform substitution switching part 16 when the signal x(m) is not the lack packet and so is the preceding packet, a signal x(m) is outputted to an output terminal 103 and a processing switching part 11. The switching part 11 executes the processing of a non-normalized processing part 19, a period detection part 18 and an amplitude correction coefficient introduction part in correspondence with the sample number (m) of the inputted packet. Thus, a signal considered to be optimum for interpolation and an amplitude correction coefficient are calculated. An amplitude correction part 14 multiplies said signal and the coefficient and outputs them to a weighing mean value arithmetic part 15. Furthermore, the prediction value is outputted to the arithmetic part 15. It calculates the interpolation signal from it and output the signal to a switching part 16. It outputs the interpolation signal which does not damage practicality.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a packet switch, and more specifically to a voice packet interpolation method for missing voice packets in a voice packet switch.

(Prior Art) In voice packet communication, real-time communication is an important element of communication quality. Therefore, if a voice packet does not arrive even after a predetermined time limit has passed, the voice packet is treated as lost, thereby guaranteeing real-time communication. Loss of voice packets also occurs, for example, when a load exceeds the processing capacity of the switch. When such a loss of audio packet occurs, the audio of that portion is not reproduced, resulting in deterioration of communication quality. Therefore, when a voice packet is lost, it is desirable to assume that the voice packet is lost and interpolate the voice packet in the missing portion to prevent deterioration in communication quality.

As an interpolation method for such missing voice packets, for example, J. Goodman, et al., Wavef.
ormSubstition Technique
s for Recovering Missing 5
peach Segments in Packet
Voice Communications IEE
E Transactions on Acoustic
s, 5peech, and Signal Pro
cessing, Volume 34, Issue 6, No. 14
There is one described on pages 40-1447 (1988, 12). Conventional technology uses pattern munching or pitch detection to transform the waveform before and after the missing part so that the connection between the non-missing part and the missing packet is smooth when interpolating the waveform to the missing part. It is shown. Specifically, add the value obtained by multiplying the waveforms immediately before and after the waveform to be interpolated to the missing part by a weighting coefficient Kn, and the value obtained by multiplying the waveforms immediately before and after the missing part by a weighting coefficient (1-Kn). do. Here, the coefficient Kn is a coefficient that approaches rOJ as it moves away from the waveform part directly interpolated into the missing part, and here a raised-cosine curve is used.

(Problems to be Solved by the Invention) However, in such conventional techniques, when deforming the front part of the missing part, the sample value input to the memory after the packet loss occurs is changed. For this reason, sample values cannot be output until it is confirmed whether or not subsequent packets are missing, resulting in a processing delay that impairs the real-time nature of voice packets.

It is an object of the present invention to provide an audio packet interpolation method that eliminates the drawbacks of the prior art and performs interpolation processing for missing packets and processing for smoothly connecting missing portions and non-missing portions without causing any processing delay. shall be.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the present invention provides an audio packet interpolation method in which a lost audio packet is interpolated into the missing part using a signal of an audio packet input before the audio packet is lost. In preparation for the loss of audio packets, the interpolation signal of the audio packet to be input next to the latest input audio packet is selected from among the audio packets input in advance as the interpolation sample value, and the interpolated signal of the audio packet input in advance is The signal of the next input audio packet is determined as a predicted sample value using a predetermined signal among the signals of When a packet is missing, an interpolation signal containing the calculated weighted average value in the average value area is interpolated to the missing part of the audio packet.

(Function) According to the present invention, when a sample of an audio packet is input, an interpolation signal of an audio packet to be input next to the latest input audio packet is selected from audio packets input in advance as an interpolation sample value. do. Furthermore, the signal of the next input audio packet is determined as a predicted sample value using a predetermined signal among the audio packet signals input in advance. Then, a weighted average value in the average value region of the interpolated signal is calculated using the predicted sample value and the interpolated sample value. When a voice packet is missing, the calculated weighted average value is assigned to the average value area, and this is used as an interpolation signal to be interpolated to the missing portion of the voice packet.

(Example) Next, an example of the audio packet interpolation method according to the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1, there is shown an embodiment in which the voice packet interpolation method according to the present invention is applied to a PCM interpolation processing device provided on the packet receiving side of a voice packet switch. The interpolation processing device l is connected to an input terminal lot for inputting a PCM signal of an audio packet and a control terminal 102 for notifying the loss of an audio packet. This is a device that outputs from an output terminal 103.

FIG. 2 shows various parameter values in this embodiment. As shown in the figure, in this embodiment, the length of the audio packet is 32 bytes, and the length of the template, which is the basic signal for detecting similar signals of missing packets, is 30 bytes) (
3,75m5), and the search window, which is the search range for detecting similar signals of the missing packet, is 128 bytes) (
1Bms). The search window length is determined by taking into account the average pitch period of the voice and the amount of calculation, and the template length is determined to be an appropriate value based on the prevention of false detection of the optimal matching waveform and the amount of calculation. Further, in this embodiment, the front weighted average area length of the missing packet is 4 bytes, and the rear weighted average area length, which is the head of the packet following the missing packet, is 8 bytes. Note that these numerical values indicate one example, and of course the present invention is not particularly limited to these values.

In addition, in this embodiment, the audio packet is 32 samples (1 byte = 1 sample), the sample number m indicates whether it is the sample of one packet, and the m-th sample value of the input packet signal is x ( m), and further converts the (n+m-1)th sample value in the search window to y(n+
m-1) respectively. In this embodiment, the first 30 samples of the most recently inputted packet are used as a template, and the four packets inputted before this are used as a search window.

Returning to FIG. 1, the waveform replacement switching section 16 connects the input terminal 101 via the input 150 and the control terminal 1 via the input 152.
02 and further connected to the weighted average value calculation unit 15 via an input 158. The switching unit 18 determines whether or not a voice packet is lost based on the LO3S signal from the control terminal 102, and performs interpolation processing when a packet is lost. That is, when the switching unit 16 determines that the input packet is a missing signal, the switching unit 16 replaces the waveform with the interpolated signal of the missing packet that has been subjected to the front weighting process input from the weighted average value calculation unit 15, and outputs 154. The signal is output to the output terminal 103 via. The switching unit 1B also stores whether or not a packet is missing, inputs a signal whose leading part is subjected to rear weighting processing to the packet following the missing packet from the calculation unit 15, and replaces the waveform with this signal. It is output to the output terminal 103. The switching unit 1B outputs the PCM signal input from the input terminal 101 as it is to the output terminal 103 if the input packet is not a missing packet or a packet following the missing packet. The output 154 is also connected to the processing switching unit 11 , and the signal output to the output terminal 103 is input to the switching unit 11 .

The processing switching unit 11 selects the sample number m of the input packet.
According to
8 and the amplitude correction coefficient deriving section 17 are executed. That is, the switching unit 11 selects the sample number m of the input template.
When m=1 to 30, the difference calculation processing unit 19
When m-31, the period detecting section 18 is executed, and when m-32, the amplitude correction coefficient deriving section 17 and the sample value predicting section 20 are executed.

The difference calculation processing unit 19 is a processing unit that performs non-standardized difference calculation processing for detecting the optimal matching waveform between the template sample and the PCM signal of the packet stored in the memory update unit 12. The arithmetic processing unit 18 calculates the formula (1
) is repeatedly executed until m=1 to 30.

D(n)=Σl x(m)−y(n+m−1)I (n
=1. ,,,,N-M+1) However, ! (In):
Each sample value M in the template: Number of samples in the template y (n+m-1) : Each sample value N in the search window N: Sample in the search window D(n) obtained by the calculation process of formula (1) is output 1
The signal is sent to the period detection section 18 via 80.

When the period detection unit 18 is activated by the processing switching unit 11,
D(n) input from the memory update unit 22, that is, D(
1) Select n that is the minimum value of D(9?). Then, the value obtained by subtracting the selected value of n from 129 is the time interval between the template and the optimal matching waveform, that is, the fundamental period PITCH of the audio waveform. The detection unit 18 outputs an output 18
The fundamental period PITCH is outputted to the buffer memory pointer designation unit 13 via the output 186 and to the amplitude correction coefficient derivation unit 17 via the output 186.

Normally, the amplitude of the envelope of a speech waveform changes, so by correcting the amplitude during waveform replacement processing and then performing waveform replacement, it can be expected that the sound will not feel unnatural.

It is conceivable that the amplitude correction coefficient KLVL used for amplitude correction takes the ratio of the sum of absolute values of the amplitudes of the template and the optimal matching waveform, as shown in equation (2), for example.

Σ1x(i) KLVL= z2(2) Σ1YII(i)l where, x(i): template sample value ylI(i)
: When the optimum matching waveform sample value coefficient deriving unit 17 in the search window is activated by the processing switching unit 11,
Sample values ylI(i), (i=1..., 32) of the optimal matching waveform in the search window corresponding to the fundamental period PITCH input from the detection unit 18 and sample values x(i), (i = L..., 3
2) Enter. The derivation unit 17 calculates an amplitude correction coefficient KLVL, which is the ratio between the power of the template part and the power of the optimal matching waveform part in the search window, and calculates the amplitude correction coefficient KLVL.
The LVL is outputted to the amplitude correction section 14 via the output 184.

Furthermore, in the case of m=32, after the coefficient deriving section 17 finishes processing, the sample value prediction section 20 is activated by the processing switching section 11. The prediction unit 20 calculates a value predicted to be input next from the sample values in the memory update unit 12. For example, the formula for calculating the predicted value is K(Ilpre) = x(
m-1) or K(mpre) = 2 X x(m-1) -x(m
-2)+ KLVLX (z(m)-2X z(m-
1)+z(m-2) ) (3) Or x(mpre) -x(m-1)+(x(m-1)-
xc addition-2))x (z(m)-z(m-1))/
(z(m-1) -z(m-2)) (4) However, x(mpre): Predicted value x(m-i) of the input sample
: Sample value i times before the input sample 2 (Il+): Sample value z(m−i)·2 (m) i times before the input sample, etc. In this way, the predicted value X(mpre) is determined based on the sample value input immediately before, so this predicted value X(mpre)
By using pre) as a parameter of the front weighted average region of the interpolation signal to be interpolated to the missing packet, it is possible to calculate an appropriate interpolation signal that matches well with the packet immediately before the missing packet.

The prediction unit 20 outputs the calculated predicted value X (mp r e ) to the weighted average value calculation unit 15 via the output 182.

Note that the switching unit 11 does not perform the above processing for the first two packets. The switching unit 11 is connected to the control terminal 102 via an input 152, from which it receives the PSYNC signal of the packet. The PSYNG signal is included in the control signal corresponding to the input sample value along with the Loss signal,
When the PSYNC signal is 1, it indicates that the input sample is the first sample of the packet. The switching unit 11 monitors whether three or more packets have been input in succession by counting the PSYNC signals. If any of the first three packets is lost, pre-interpolation is performed since the initial value of the basic period PITCH is 32. The processing switching section 11 is connected to the buffer memory updating section 12 via an output 160 and outputs the signal X (m) inputted from the waveform replacement switching section 16 to the memory updating section 12 .

The buffer memory update unit 12 is a storage unit that can store a search window for 4 packets, a template for 1 packet, and the latest input sample for 1 sample, and stores PCM signals of, for example, 5 packets and an audio packet for 1 sample. It is.

The update unit 12 updates the sample value every l sample input, and when the storage capacity is exceeded, the update unit 12 updates the sample value in the order of input.
Discard M signal.

The eight-point memory pointer designation unit 13 searches the memory update unit 12 for a PCM signal for interpolating the missing audio packet based on the basic cycle PITCH input from the synchronization detection unit 18, and uses this as a signal x (mOpt). It is output to the amplitude correction section 14.

The amplitude correction unit 14 inputs the signal x (mopt) and the correction coefficient KLVL, and multiplies these input signals to generate a compensated signal KLV that is subjected to amplitude correction according to the audio waveform envelope.
Calculate LX
Weighted average value calculation unit 15 as X x (mopt)
Output to.

The weighted average value calculation unit 15 calculates the interpolation signal KLVLX x
(mopt) and the predicted value x (mpre) are respectively input, and the interpolated value containing the signal value weighted in the front weighted average region by Equation (5), that is, the interpolated signal x (mlnew) of the missing packet is obtained. Calculate. The calculation unit 15 is also connected to an input terminal lot via an input 150, from which the signal x(m) is input.

The calculation unit 15 calculates an interpolated value including a signal value obtained by weighting the rear weighted average region using, for example, equation (6), that is, an interpolated signal x (m2new) following the missing packet.

x(mlnew)−+ (i) X x(mpre)
+ (+ (+) to 1-)XKLVLXx(mop
t) (i=1.,,,s+)KLVLXx(mo
pt) (i=s++1.,,.32)(+=1.
−、−． 81) Sl: Front weighted average region length.

x (mopt): Sample value x (+
o2new) = K2(i) Xi(m)+ (+-
2(i) )XKLVLXx(m(,ptp) (
i=1. ,,,,s2)x(m) (i=s
2+1. ,,,． 32) (i=1.,..., s2) S2: Rear weighted average value region length x (moptp): Sample value one pitch period before in the previous packet section. Note that in this embodiment, as described above, 31=4 samples and 52=8 samples. That is, the interpolated signal x
(mlnew) is the interpolated signal x (mlnew) when the sample number i is in the range of 1 to 4, x(mlnew) = Kl
(i) X x (mpre) + (1-to + (i
) )X KLVLX
lne++)=KLVLX x(mopt).

Similarly, the interpolated signal x(m2new) is expressed as x(m2new)-2(i) Xi(m)+(1-
2(i) ) (m
2new)-x(m). The average calculation unit 15 calculates the calculated interpolation signals x (mlnew) and x (m2new).
is output to the waveform replacement switching unit 16 via the output 158.

Explain the operation. When the PCM signal X (m), which is an audio packet, is input from the input terminal 101 to the input 150 to the waveform replacement switching unit 16, the switching unit 16 determines whether or not this signal is a missing packet based on the LO3S signal. If it is not a packet, check whether the packet input immediately before is a missing packet. When the signal x(m) is not a missing packet and the previous packet is not a missing packet, this input signal x(m) is outputted by the switching section 16 to the output terminal 103 and the processing switching section 11.

When the signal x (m) is input to the processing switching unit 11, the processing switching unit 11 determines that the sample number m of the signal x (m) is 1 to
This signal at 30! (+I+) as a template, the difference calculation processing unit 19, when the sample number m is 31, the period detection unit 18, and when the sample number m is 32, the amplitude correction coefficient derivation unit 17 and the sample value prediction unit 2.
Execute 0. As a result, the fundamental period PIT of the audio waveform, which is the time interval between the template and the optimal matching waveform,
CH is calculated, and when the input packet next to the packet containing the signal x(m) is a missing packet, the signal x(11opt
) is selected from within the search window stored in the buffer memory updating unit 12. Also, this signal x(IIlo
pt) is suitable for the envelope of the audio waveform.
LVL is calculated.

When the signal x (lopt) and the correction coefficient KLVL are input to the amplitude correction section 14, the correction section 14 converts the signal x (m,
p t ) by the amplitude correction coefficient KLVL, and the resulting signal KLVLX x (mopt) is weighted and average value calculation unit 1
Output to 5. Further, when the sample number m is 32, the sample value prediction section 20 is executed by the switching section 11. As a result, the prediction unit 20 uses the sample value in the memory update unit 12 to determine the signal! The signal of the next input packet after (32) is calculated as a front weighting parameter, and this is output to the average value calculation unit 15 as a predicted value x (mpre).

Signal KLVLX x(mopt) and predicted value X(”P
re) is input to the arithmetic unit 15, the arithmetic unit 15 calculates an interpolated signal x(ml
new) and outputs it to the waveform replacement switching unit 16. The interpolated signal x (mlnew) is output to the switching unit 16 almost at the same time that the switching unit 18 confirms that the input signal is a missing packet.
Therefore, even if a packet is lost, the switching unit 16 uses this signal x (mlnew) to send the interpolated signal
It can be output to 03. Also, the signal z(mines
), the 4 bytes of the front weighted average area are weighted, so the connection with the previously input signal x(m) becomes smooth.

When the waveform replacement switching unit 16 inputs the signal x (m) included in the non-missing packet following the missing packet, the replacement switching unit 16 inputs the signal x (m) from the weighted average value calculation unit 15 instead of this signal x (m). The interpolated signal x (m2nei+) is output to the output terminal 103. When the signal x(m) included in the non-missing packet following the missing packet is input to the average value calculation unit 15, the calculation unit 15 calculates the signal x(m) and the sample value one pitch period before in the previous missing packet part. X(Illopt
p), and interpolation signal x(m2ne
w) is calculated. The values of the interpolated signal x (m
2new), weighting is applied to the part connected to the missing packet, that is, the 8-byte rear weighted average value area, so that the missing packet and the next interpolated signal x(m2new) are weighted.
The connection with w) becomes smooth.

As described above, according to this embodiment, by using a new interpolation value of the weighted average value of the predicted value and the interpolation value, even when an audio packet is missing, an appropriate PCM signal that does not give an unnatural feeling can be interpolated into the missing part. is possible. Furthermore, in this embodiment, the audio waveform to be interpolated is always determined at the timing when the first sample of a packet is input, so even if a packet is lost, interpolation can be performed without impairing the real-time nature of the audio packet.

In this embodiment, the present invention is applied to a PCM interpolation processing device that interpolates missing audio packets, but the present invention is not necessarily limited to this. For example, the present invention is applied to the receiving side of a transmission device that transmits image information etc. , or a digital audio signal reproducing device.

(Effects of the Invention) As described above, according to the present invention, a method is introduced in which the weighted average value of the predicted value and the interpolated value is used as the sample value of the connection portion before and after the missing packet. Therefore, even if a missing packet occurs, it is possible to interpolate the missing packet and smoothly connect the missing part and the non-missing part of the packet without impairing the real-time nature of the packet.

[Brief explanation of the drawing]

FIG. 1 is a functional block diagram in which an embodiment of the audio packet interpolation method according to the present invention is applied to a PCM interpolation processing device. FIG. 2 is a parameter configuration diagram showing the configuration of various parameters used in FIG. Explanation of symbols of main parts PCM interpolation processing device Processing switching section Buffer memory updating section Buffer memory pointer specifying section Amplitude correction section Weighted average value calculation section Waveform replacement switching section Amplitude correction coefficient derivation section Period detection section Non-standardized difference calculation processing section Sample value prediction part

Claims (1)

[Claims] In an audio packet interpolation method that interpolates a missing audio packet to the missing part using a signal of an audio packet input before the audio packet is missing, the method is configured to interpolate a missing audio packet into the missing part using a signal of an audio packet input before the audio packet is lost. The interpolated signal of the audio packet to be input next to the audio packet is selected from among the audio packets input in advance as an interpolation sample value, and the next A signal of an input audio packet is obtained as a predicted sample value, and a weighted average value in the average value region of the interpolated signal is calculated using the predicted sample value and the interpolated sample value. A voice packet interpolation method characterized in that an interpolation signal containing the calculated weighted average value in the average value region is interpolated into the missing portion of the voice packet.