JPH1078791A - Pitch converter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a pitch converter capable of independently adding a change of formant and a change of pitch to speech waveforms. SOLUTION: A first pitch conversion section 10 segments phoneme waveforms from speech waveforms and repetitively outputs these phoneme waveforms at the period corresponding to a first pitch conversion ratio S1 . A second pitch conversion section 20 stores the speech waveform obtd. from the first pitch conversion section 10 into a memory, reads the speech waveforms out of this memory at the reading-out speed corresponding to a second pitch conversion ratio S2 out of the memory and outputs these waveforms. A controller 30 outputs the first pitch conversion ratio S1 and the second pitch conversion ratio S2 in correspondence to the desired pitch conversion ratio S0 and the conversion ratio F0 of a formant frequency.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a pitch converter used for converting a pitch of an audio signal.

[0002]

2. Description of the Related Art In the field of handling audio signals, pitch converters for raising and lowering the pitch of audio signals are often used.
For example, in recent karaoke apparatuses, by using such a pitch converter, a function of changing the pitch of a singing voice input via a microphone and generating a harmony sound that forms a consonant with the singing voice. Many are equipped with.

Conventionally, such a pitch converter performs pitch conversion by expanding and contracting a speech waveform to be processed in the time axis direction, and performs pitch conversion by cutting out and resynthesizing a phoneme waveform from the speech waveform. There were two types of things. The outline of these pitch converters will be described below.

A. Pitch converter for performing pitch conversion by expanding and contracting an audio waveform in the time axis direction FIG. 2 shows an example of the configuration of this pitch converter.
This pitch converter comprises a memory 1, multipliers 2 and 3, and an adder 4, as shown.

[0005] The pitch converter is supplied with time-series sample data of the audio waveform to be processed at a fixed sample rate, and the sample data thus supplied is sequentially written into the memory 1. Then, in parallel with the writing of the sample data, reading of the already written sample data is performed. Here, the reading of the sample data is performed at a speed obtained by multiplying the writing speed by a desired pitch conversion ratio (the ratio of the pitch of the converted speech waveform to the pitch of the converted speech waveform). Therefore, the sample data representing the voice waveform obtained by compressing or expanding the original voice waveform in the time axis direction according to the pitch conversion ratio is read from the memory 1. This pitch converter basically performs pitch conversion of a speech waveform based on such a principle.

The duration of a speech waveform to be processed is relatively long, and the time-series sample data generally has a huge data length. For this reason, a method of using the memory 1 as a ring buffer, that is, a method of cyclically and repeatedly writing sample data to a storage area in the memory 1 and sequentially extracting the sample data in a manner following this write operation is generally adopted. It is a target.

However, in the pitch conversion operation,
Since the sample data is read at a reading speed different from the writing speed, the difference between the writing address and the reading address gradually increases. If the period for writing and reading sample data is longer than a certain limit, even if an attempt is made to read certain sample data from the memory 1, the sample data has already been rewritten by another sample data at that point. As a result, a state occurs in which reading cannot be performed.

In order to avoid such inconveniences, the pitch converter shown in FIG. 2 performs an address control for making the read address coincide with the write address at regular intervals, and performs sample control in the memory 1 under this address control. Data is being read. Here, this operation will be described with reference to FIG.

First, FIG. 3A shows an input voice waveform corresponding to the sample data written in the memory 1. The waveform W ₁ shown in FIG. 3 (b) shows the voice waveform corresponding to sample data read from the memory 1 under the address control. In this example, the sample data of the original audio waveform (see FIG. 3A) is read from the memory 1 at a slower speed than at the time of writing, and the original audio waveform is expanded in the time axis direction. 3 (b) is read from the memory 1. Then, each sample data of the section A ₁ in FIG. 3 (a) corresponds to each sample data of the section A ₁ 'in FIG. 3 (b). In this example, the section A ₁
Control to match the read address to the write address is performed when the read is performed for the sample data corresponding (※ mark), writing and reading of sample data corresponding to the next segment A ₂ it has been made to.

By the way, when the control for matching the read address to the write address is performed periodically, a discontinuity point occurs in the audio waveform read from the memory 1 at the time when the control is performed. Therefore, FIG.
In the pitch converter shown in (1), sample data of a smooth speech waveform with no noticeable discontinuities is generated by the method described below.

First, in the pitch converter shown in FIG. 2, the reading of the sample data is performed in parallel with the reading of the sample data. In the reading of this sample data, the control of matching the read address with the write address is repeated at a timing shifted by a half cycle from the address control described above (*
mark). Waveform W ₂ shown in FIG. 3 (c) shows a speech waveform corresponding to sample data read from the memory 1 in this way.

The sample data of the audio waveforms W ₁ and W ₂ at which discontinuities appear at timings shifted from each other by a half cycle are read in parallel from the memory 1 and supplied to the multipliers 2 and 3. You. And each of the multipliers 2 and 3
Accordingly, each sample data is multiplied by a multiplication coefficient, and each multiplication result is added by the adder 4 and output as final sample data. At this time, control is performed to continuously and smoothly change the multiplication coefficients of the multipliers 2 and 3 so that the sample data near the discontinuous point is multiplied by 0. This is a so-called crossfade.
By this operation, sample data of a voice waveform having a desired pitch and inconspicuous discontinuities can be obtained.

B. Pitch converter for extracting and resynthesizing a phoneme waveform from a speech waveform FIG. 4 shows a configuration example of this pitch converter. This pitch converter includes a waveform cutout unit 11 and a re-synthesis unit 12.
It consists of.

The time series sample data of the audio waveform to be processed is taken into the waveform extracting section 11, and the sample data of the phoneme waveform is extracted. The extraction of the sample data of the phoneme waveform can be performed by, for example, a process of detecting the pitch of the audio waveform and multiplying the sample waveform by the Hanning window having a length of two pitch periods.

The sample data sequence of the phoneme waveform obtained in this manner is repeatedly output by the resynthesizer 12. The output of the sample data sequence of the phoneme waveform is performed at a pitch cycle obtained by multiplying the pitch of the original speech waveform by the reciprocal of the required pitch conversion ratio. As a result, sample data of the pitch-converted audio waveform is obtained.

[0016]

By the way, the above-mentioned 2
According to the pitch converter that performs pitch conversion by expanding and contracting the audio waveform in the time axis direction among the pitch converters of the type, the formant frequency of the audio waveform is increased or decreased at the same ratio as the pitch. Will be changed. On the other hand, a pitch converter that extracts a phoneme waveform from a speech waveform and resynthesizes the speech waveform does not change the formant of the speech waveform, but changes only the pitch.

An object of the present invention is to provide a pitch converter that can independently change the formant and the pitch of an audio waveform by utilizing the features of the pitch converter described above.

[0018]

According to the present invention, there is provided a first pitch conversion means for extracting a phoneme waveform from a speech waveform and repeatedly outputting the phoneme waveform at a cycle corresponding to a first control signal; A second pitch conversion means connected to an input side or an output side of the pitch conversion means, the second pitch conversion means for expanding and contracting an audio waveform in a time axis direction at a ratio corresponding to the second control signal, and outputting A gist of the invention is a pitch converter comprising: control means for outputting the first control signal and the second control signal in accordance with the pitch conversion ratio and the conversion ratio of the formant frequency.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments will be described to make the present invention easier to understand. Such an embodiment shows one aspect of the present invention, and does not limit the present invention, and can be arbitrarily changed within the scope of the present invention.

FIG. 1 is a block diagram showing a configuration of a pitch converter according to an embodiment of the present invention. As shown, the pitch converter includes a first pitch converter 10 and a second pitch converter 20 that are cascaded, and a controller 30 that controls these pitch converters.

Here, the first pitch conversion unit 10 is a means for performing pitch conversion by extracting and resynthesizing a phoneme waveform from a speech waveform, and has the same configuration as that already described with reference to FIG. Have. The second pitch conversion unit 20 is means for performing pitch conversion by expanding and contracting the audio waveform in the time axis direction, and has the same configuration as that already described with reference to FIG. The sample data of the audio waveform input from the input terminal 31 is subjected to pitch conversion in two stages by the first pitch conversion unit 10 and the second pitch conversion unit 20, and is output from the output terminal 32. You.

The pitch conversion ratio S ₀ and the conversion ratio of the formant frequency (the ratio of the formant frequency of the converted sound waveform to the formant frequency of the converted sound waveform) F are transmitted to the controller 30 by the operation of operation means (not shown).
₀ is given. The controller 30 calculates the pitch conversion ratios S ₁ and S ₂ based on these pieces of information, and supplies them to the first pitch conversion unit 10 and the second pitch conversion unit 20, respectively. The details are as follows.

First, in the pitch conversion by the second pitch conversion unit 20, the conversion of the formant frequency is performed at the same ratio as the pitch conversion ratio. Therefore, the controller 3
0 supplies the pitch conversion ratio S ₂ of the same value as the conversion ratio F ₀ of the formant frequency given from the operating unit to the second pitch conversion unit 20.

Next, in order to set the pitch conversion ratio of the entire pitch converter to S ₀ , the pitch conversion ratio S ₀ is divided by the pitch conversion ratio S ₂ supplied to the second pitch conversion unit 20, and the division is performed. The pitch conversion ratio S ₁ having the same value as the result S ₀ / S ₂
To the first pitch converter 10.

[0025] As a result, the sample data of the speech waveform inputted from the input terminal 31, the first pitch converter 10, is converted to S ₁ times the pitch while maintaining the formants of the current situation. Then, the first pitch converter 10
To the sample data of the voice waveform obtained from the pitch conversion is performed by the second pitch converter 20, both pitch and formant frequencies are changed _twice S.
As a result, sample data of an audio waveform having a pitch S ₀ times the original audio waveform and a formant frequency changed to the original F ₀ times is output to the output terminal 32.

As described above, in the present embodiment, the first pitch converter 10 that performs only the pitch conversion while maintaining the formant, and the second pitch converter 20 that changes both the formant and the pitch, Since the pitch conversion is performed in two stages, the change of the pitch and the change of the formant can be independently applied to the input voice waveform.

In the above-described embodiment, the first pitch converter 10 is arranged at the front stage and the second pitch converter 20 is arranged at the rear stage. May be arranged in the first stage, and the first pitch conversion unit 10 may be arranged in the second stage.

[0028]

As described above, according to the pitch converter of the present invention, there is an effect that a formant change and a pitch change can be independently applied to a voice waveform.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a pitch converter according to an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration example of a conventional pitch converter.

FIG. 3 is a waveform chart showing an operation of the pitch converter.

FIG. 4 is a block diagram showing another configuration example of a conventional pitch converter.

[Explanation of symbols]

10 first pitch converter, 20 second pitch converter, 30 controller.

Claims (1)

[Claims] 1. A first pitch conversion means for cutting out a phoneme waveform from an audio waveform and repeatedly outputting the phoneme waveform at a cycle corresponding to a first control signal; and an input side or an output of the first pitch conversion means. A second pitch conversion means for expanding and contracting the audio waveform in the time axis direction at a ratio corresponding to the second control signal and outputting the same, and converting the desired pitch conversion ratio and formant frequency Control means for outputting the first control signal and the second control signal in accordance with the ratio.