JPH08286696A - Speech speed conversion and decoding method - Google Patents

Abstract

PURPOSE: To provide a speech speed conversion and decoding method which can convert and reproduce a speech speed by adding only slight quantity of operation. CONSTITUTION: By using a voice decoding method reproducing a voice signal in which a voice signal is coded using a voice coding method including a parameter relating to a pitch period in a coding data and it is recorded in a memory 1 while it is decoded, a voice signal is compressed and expanded by interpolating and thinning waveforms of a decoded voice signal in a time base of the voice signal by a speech speed conversion processing means 5 based on a pitch period of a parameter relating to a pitch period being a coding data, and a speech speed conversion is performed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to speech rate conversion used in an answering machine or the like, for example, when a voice signal is encoded and recorded in a solid-state memory and is decoded, the pitch of the voice (pitch of voice). The present invention relates to a speech speed conversion / decoding method for performing so-called fast listening and slow listening in which the speech speed (voice speed) is changed without changing the output.

[0002]

2. Description of the Related Art Conventionally, as a decoding method for reproducing a voice by changing the speech speed without changing the pitch of the voice used in this type of voice signal encoding / decoding system, the conventional method shown in FIG. Those based on devices to which the method is applied are known. In the figure, 1 is a memory such as a solid-state memory for storing an encoded audio signal, 2 is a decoding processing means for decoding the encoded audio signal read from the memory 1, and 3 is a temporary decoded audio signal. A buffer for storing data 4 in order to calculate a pitch cycle from the decoded speech signal read from the buffer 3, and 5 in accordance with the pitch cycle calculated by the pitch cycle calculating means 4 based on the pitch cycle. It is a voice speed conversion processing means for converting the voice speed of the voice and outputs a voice signal whose voice speed has been converted.

In this method, the pitch period required for speech speed conversion is calculated by the pitch period calculating means 4 from the decoded speech signal read from the buffer 3 by an estimation processing method such as an autocorrelation method or a cepstrum method. Overlap addition method by pointer control using the calculated pitch period
(PICOLA: Pointer Interval Control OverlapAnd
Add), time-domain harmonic structure expansion / contraction method (TDHS: Time-Domai)
n Harmonic Scaling) etc., the speech speed is converted by the speech speed conversion processing means 5 by a method of converting the speech speed by inserting or thinning the speech waveform on the time axis of the speech signal. In this way, the desired speech speed conversion can be performed by newly calculating the pitch period from the decoded voice signal.

[0004]

However, in this conventional speech speed conversion method, a new problem is that the pitch period is newly calculated, and in order to do this, the amount of calculation and the instruction ROM increase. Especially, when the compression rate of the audio signal becomes high (when the speech speed is increased), the decoding process must be performed three times or more as much as usual, which causes a problem that the process cannot catch up.

The present invention solves the above-mentioned conventional problems, and provides a speech speed conversion / decoding method capable of converting the speech speed for reproduction by adding a small amount of calculation.

[0006]

In order to achieve the above object, the present invention encodes a voice signal using a voice encoding method in which encoded data includes a parameter related to a pitch period and records it in a memory or the like. And a voice decoding method for decoding it to reproduce a voice signal, and decoding the decoded voice signal based on the pitch period indicated by the parameter related to the pitch period which is the encoded data, and the time axis of the voice signal. In the above, the waveform is inserted and thinned to compress and expand the voice signal to convert the speech speed.

[0007]

In the conventional method, the pitch period of the speech speed conversion for compressing and expanding the voice signal on the time axis is controlled by using the pitch period calculated by the autocorrelation method, the cepstrum method, etc. Although the conversion is performed, usually, the speech speed conversion of the format in which the audio signal is compressed and expanded on the time axis described above causes a large deterioration in quality even if the pitch period to be controlled is slightly different. Therefore, in the present invention, the speech period conversion / decoding can be performed without increasing the calculation amount by using the pitch period of the frame unit such as 20 ms, which is usually obtained by the encoding / decoding method of the voice signal. It can be carried out.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIG. It should be noted that the same parts as those of the above-mentioned conventional one are denoted by the same reference numerals. In the figure, 1 is a memory such as a solid-state memory for storing an audio signal encoded by an encoding method having a parameter related to the pitch period such as a pitch period or a lag in the encoded signal, and 2 is a code read from the memory 1. Decoding processing means for decoding the deciphered speech signal, 3 is a RAM for temporarily storing the deciphered speech signal and parameters relating to the pitch period used at the time of decoding.
And the like, the buffer 5 is a voice speed conversion processing means for converting the voice speed of the decoded voice signal based on the pitch period represented by the parameter related to the pitch period, and outputs the voice signal whose voice speed is converted. To do.

Next, the operation will be described. First, TC-WVQ (Transform Coding Wi
th Weighted Vector Quantization) method, RPE-LT
P (Regular Pulse Excitation-Long Term Predictio
n) method, VSELP (Vec
tor Sum Excited Linear Prediction) method, such as a ROM in which a voice signal is encoded and stored in advance by an encoding method that has a parameter related to the pitch period such as pitch period or lag in the encoded signal, RAM that is temporarily stored, etc. The coded audio signal read from the memory 1 including the decoding processing means 2
The decoded voice signal and the parameters related to the pitch period decoded at that time are temporarily stored in the buffer 3 for a plurality of frames. Then, the speech speed conversion processing means 5 uses the voice signal stored in the buffer 3 and the pitch period of the parameter related to the pitch period used when the voice signal is decoded, by the pointer control overlapping addition method. (PICOLA), time domain harmonic structure expansion / contraction method (TDHS), etc., the speed of the reproduced voice is reduced by decimating the audio signal on the time axis in pitch waveform units or inserting the same or similar waveform. Adjust to any speed.

By using the speech signal thus decoded in the decoding processing and the parameter relating to the pitch period used in decoding the speech signal in the speech speed conversion processing, the reproduction speed of the decoded speech is reproduced. Can be adjusted to any speed.

[0011]

As is apparent from the above embodiments, the present invention converts the voice signal decoded by the decoding process and the voice speed conversion parameter related to the pitch period used when decoding the voice signal. Since it is not necessary to newly calculate the pitch period for the voice speed conversion processing by using it for the processing, the reproduction speed of the decoded voice can be adjusted to an arbitrary speed without increasing the calculation amount and the program size. .

[Brief description of drawings]

FIG. 1 is a circuit block diagram showing an example of an apparatus that can implement a speech speed conversion decoding method of the present invention.

FIG. 2 is a circuit block diagram showing an example of an apparatus that can implement a conventional speech speed conversion / decoding method.

[Explanation of symbols]

1 ... Memory, 2 ... Decoding processing means, 3 ... Buffer,
4 ... Pitch cycle calculation means, 5 ... Speech speed conversion processing means.

Claims (1)

[Claims] 1. A voice decoding method in which a voice signal is encoded by using a voice encoding method in which encoded data includes a parameter related to a pitch period, recorded in a memory or the like, and decoded to reproduce the voice signal. Based on the pitch period indicated by the parameter related to the pitch period which is the encoded data, the decoded voice signal is inserted and thinned out on the time axis of the voice signal to compress and expand the voice signal. Then, the speech speed conversion decoding method is characterized by performing speech speed conversion.