JPS63108400A - Voice encoder - Google Patents

Abstract

(57) [Abstract] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a multi-pulse driven audio encoding system.

In particular, the present invention relates to means for searching for pulses with excellent encoded sound quality under a limited amount of calculation.

〔overview〕

The present invention provides a speech encoding method in which a predetermined number of pulses within the frame are obtained by searching outside the cross-correlation frame, and when the number of pulses outside the cross-correlation frame reaches a predetermined number, the absolute value of the cross-correlation is determined. By limiting the search for the maximum value within the cross-correlation frame, it is possible to improve the encoded sound quality with a limited number of pulse search loops.

[Conventional technology]

One of the methods for encoding audio at a bit rate of 8 to 16 kbps is 10 bits per frame of about 20 milliseconds.
There is a multi-pulse driven speech encoding method that drives a linear predictive filter with ~30 pulses. In this method, it is necessary to find pulses so that the error between the human signal and the composite signal is minimized, but as a method to simplify the calculation, the absolute value of the cross-correlation between the input signal and the impulse response of the linear prediction filter is maximized. A method has been proposed in which a pulse is set at the position where the cross-correlation pulse is set, and the second and subsequent pulses are obtained by subtracting the impulse response from the position at which the cross-correlation pulse was set, and then obtaining the new cross-correlation in the same way.

In addition, in order to remove the influence of the previous frame from the cross-correlation between the input signal and impulse response, the input of the synthesis filter is set to zero, the influence signal is obtained and subtracted from the input signal, and the noise is masked using the auditory characteristics. For auditory weighting, it is common to obtain a signal passed through a filter and use the cross-correlation between this signal and the impulse response. Pulse search is characterized by searching not only within the cross-correlation frame but also outside the frame, and finding a predetermined number of pulses within the frame provides better encoded sound quality than searching only within the frame. It is known to do.

The thus searched multipulses are applied to multipulse driven speech encoding means.

[References]

Tokai University Press series Furui; Digital audio processing 1) 5~
1) 6 pages.

[Problem that the invention seeks to solve]

However, in order to perform multi-pulse encoding in real time, it is necessary to limit the overall number of pulse search loops, and in that case, it is not always necessary to perform pulse search outside the frame for better encoded sound quality. was not limited.

Even if the number of pulse search loops is limited,
It is an object of the present invention to provide a speech encoding device that can realize excellent encoded sound quality.

[Means for solving problems]

The present invention provides means for determining linear prediction coefficients by linear predictive analysis of an audio input signal, means for determining an impulse response of a linear prediction filter corresponding to the linear prediction coefficient, and cross-correlation between the audio input signal and the impulse response. a first pulse located at a time position where the absolute value of this cross-correlation is maximum and whose amplitude is the magnitude of this cross-correlation; and an amplitude determined based on the cross-correlation within a predetermined range. and a means for generating a predetermined number of pulses following the first pulse, wherein the number of pulses outside the predetermined range of the cross-correlation is The present invention is characterized by comprising means for limiting the search for the maximum value of the absolute value of the cross-correlation within the predetermined range of the cross-correlation when the predetermined number is reached.

[Effect]

A first pulse with the magnitude of the cross-correlation is placed at the position of the maximum absolute value of the cross-correlation between the impulse response of the linear prediction filter corresponding to the linear prediction coefficient determined by linear prediction analysis of the audio input signal and the audio input signal. stand up The autocorrelation of the impulse response is normalized to the pulse size at the position where the cross-correlation pulse is set, and this value is subtracted from the cross-correlation to obtain a new cross-correlation within the predetermined cross-correlation range. When generating a predetermined number of pulses,
A pulse search is performed not only within the cross-correlation range but also outside the cross-correlation range, and when the number of pulses outside the cross-correlation range reaches a predetermined number, search for the maximum absolute value of the cross-correlation. is limited to within a predetermined range of cross-correlation.

〔Example〕

Embodiments of the present invention will be described below based on the drawings.

FIG. 1 is a block diagram showing the configuration of this embodiment. This embodiment device includes a synthesis filter 1 that synthesizes the influence signal X, a weighting filter 2 that performs auditory weighting on the difference between the input signal X and the influence signal A linear prediction analysis circuit 3 analyzes and generates a linear prediction coefficient A, an impulse response generation circuit 4 calculates an impulse response R of a linear prediction filter using the linear prediction coefficient A as a coefficient, and weighting is performed by filter output X%1 Cross-correlation generation circuit 5 for calculating the cross-correlation Φ×r between Φ×r and the impulse response
, an autocorrelation generation circuit 6 that calculates the autocorrelation Φrr of the impulse response, a pulse search circuit 7 that performs a pulse search based on the cross-correlation Φxr and the autocorrelation Φrr, and encodes the output of the pulse search circuit 7 and the linear prediction coefficient A. and an encoding circuit 8 for encoding.

Next, the operation of this embodiment will be explained based on FIGS. 1 and 2. In order to remove the influence of the previous frame from the input signal X, the influence signal X is created by setting the input to zero while retaining the internal data of the synthesis filter.

is obtained, subtracted from the input signal X, and subjected to aural weighting to obtain a signal Xw that is passed through a filter 2 so as to mask noise using the audible characteristics. Perform linear predictive analysis on the input signal X,
Find the linear prediction coefficient A. The impulse response R of the linear prediction filter having the linear prediction coefficient A as a coefficient is determined, and the weighting is the cross-correlation Φxr of the signal X- and the impulse response R, and the autocorrelation Φrr of the impulse response R is determined. , cross-correlation φxr (i) of length lx, i =
Find the maximum value φxr (is) of 1 to I x, and find the maximum value φxr (is) of
A pulse of magnitude Φxr(is,) is generated at m. The autocorrelation Φrr of length ±Tr is subtracted from the cross-correlation Φxr at the position is with the magnitude adjusted to Φxr(iII+).

(i--1r to Ir) When encoding a predetermined number of pulses in this manner, for example, at 9.6 kbps, approximately 15 pulses are obtained. Since pulses may appear at the same position, in order to find pulses at 15 different positions, pulse search calculations must be performed several dozen times. The new pulse size is the sum of the previous pulse size and the current size.

As mentioned above, if audio is divided into frames of, for example, every 20 milliseconds and 15 pulses are to be found within each frame, the cross-correlation length should be longer than the frame length, and pulses will be generated not only within the frame but also outside the frame. Searching for a predetermined number of pulses within a frame results in better encoded sound quality than searching only within the frame. However, in order to perform multi-pulse encoding in real time, it is necessary to limit the overall number of pulse search loops, and in that case, it is not necessary to perform pulse search outside the frame for better encoded sound quality. was not limited.

FIG. 2 is a flowchart of pulse search when the pulse search loop is limited. Conventionally, when there is no broken line in the figure and one frame has 160 samples,
Extend the pulse search outside the frame by 40 samples to 200
If you want to perform sampling, first set the pulse search range to 2.
00 and pulse search is performed until the pulse reaches a predetermined value, 15 in FIG. 2, or until the total number of loops reaches a predetermined value, 25 in FIG. If a pulse outside the frame is not to be searched, the pulse search range is first set to 160.

The present invention is characterized in that the portion indicated by the broken line in FIG. 2 is added. That is, initially, the pulse search length is set to 200 samples including outside the frame, and when the number of pulses outside the frame reaches a predetermined number (5 in FIG. 2), the pulse search length is limited to 160 samples, which is the frame length.

〔Effect of the invention〕

As explained above, the present invention limits the pulse search range to within the frame when the number of pulses standing outside the frame exceeds a predetermined value. It has the effect of improving the sound quality.

[Brief explanation of the drawing]

FIG. 1 is a block configuration diagram showing the configuration of an embodiment of the present invention. FIG. 2 is a flowchart showing the operation of the embodiment of the present invention. 1... Synthesis filter, 2... Weighting is a filter, 3
...Linear prediction analysis circuit, 4. Impulse response generation circuit, 5. Cross correlation generation circuit, 6. Autocorrelation generation circuit, 7. Pulse search circuit, 8. Encoding circuit. . Patent applicant: NEC Corporation, -1, agent: Nao Takashi Ide0, PA': Structure of the embodiment Fig. 1

Claims (1)

[Claims] (1) Means for determining linear prediction coefficients by performing linear predictive analysis on an audio input signal; Means for determining an impulse response of a linear prediction filter corresponding to the linear prediction coefficient; and determining cross-correlation between the audio input signal and the impulse response. means, a first pulse located at a time position where the absolute value of this cross-correlation is maximum and having an amplitude equal to the magnitude of this cross-correlation, and an amplitude and time determined based on the cross-correlation within a predetermined range; and means for generating a predetermined number of pulses following the first pulse, wherein the number of pulses outside the predetermined range of the cross-correlation is When the predetermined number is reached,
A speech encoding device characterized by comprising means for limiting the search for the maximum value of the absolute value of the cross-correlation within a predetermined range of the cross-correlation.