JPH0481197B2 - - Google Patents

Description

[Detailed description of the invention] The present invention relates to a pitch adjustment device.

Pitch (sound pitch) is used when reproducing the digital signal (a sequence of sample values that are discrete with respect to time) obtained by sampling the waveform of voice or musical tone (sampling period is T _s ) and performing AD conversion. There are cases where you want to change the output. The simplest method is to perform DA conversion at each sampling period (denoted as T _s ″), which is different from the sampling time, but this is not appropriate when you want to keep the sampling period of the system fixed.In such cases To do this, it is necessary to adjust the pitch by digital signal processing.To do this, a digital signal is generated as if it were sampled at a different sampling period (T _s ), and then the pitch _is It is sufficient to output it every cycle.

Conventionally, there is a method using linear interpolation as a means for obtaining such a signal. In other words, assuming a temporally continuous signal in which discrete sample values vary linearly with respect to time, sample values are generated by sampling this signal with a period of T _s ′. . This conventional example has the advantage that the amount of calculation is small and the range of pitch change can be selected almost continuously, but the error is large and depending on the shape of the spectrum envelope of the original waveform, the range of pitch change may be within a few percent. However, the drawback was that the sound quality deteriorated considerably.

Another known conventional example is one in which interpolation is performed not by a straight line but by a function given by the following equation.

h(t)=sin(πt/ _Ts )/πt/ _Ts ) (1) Here, Ts is the sampling period described above.
In this conventional example, 1 between the sample values of the original waveform to be interpolated (x(nT _s ) and x(n+1)T _s ))
The sample (denoted as y(kT _s ')) is obtained using the following equation.

y(kT _s ′)= _L2 〓 〓 ^l=-L ₁ k(hT _s ′−(n+l)T _s )+x((n+l)T _s )
(2) In this conventional example, in order to reduce the error, y
In order to obtain one sample of (kTs'), many sample values of the original waveform must be used and many calculations are required. In addition, in order to make the range of change in pitch continuous, the value of h(t) must be calculated each time, which has the disadvantage that more calculations are required.

A third conventional example uses a digital filter. This interpolates sample values whose value is zero between the sample values of the original waveform by M-1 samples (as a result, the sampling period is T _s /M
), this sample value sequence is input to a digital low-pass filter, and its output is taken out every L samples. As a result, the sampling period becomes T _s ×L/M. In this conventional example, sound quality deterioration can be prevented by using a filter with abrupt cutoff characteristics, but in order to keep the pitch change range within a few percent, L and M must be set to several tens to 100. The disadvantage was that it required several calculations.

SUMMARY OF THE INVENTION An object of the present invention is to provide a pitch adjustment device that can be realized with a relatively small amount of calculations, has little deterioration in sound quality due to the influence of errors, and can select pitch change range almost continuously.

The present invention generates a sample value for each second period by interpolating between the sample values of the waveform sampled in the first period, and outputs the sample value in the first period. In a pitch adjustment device of the type that changes the pitch, there is a means for shortening the sampling period to one integer by interpolating a sample value of zero between the sample values of the original waveform, and a filter for the output sample value sequence. Among the output sample values of the digital low-pass filter and the digital low-pass filter, a new sample value is generated by linear interpolation between the sample values before and after the time in each second cycle, and is output in the first cycle. It is characterized by comprising means for.

According to the present invention, the amount of calculation is small, the range of change in pitch can be selected almost continuously, and at the same time, deterioration in sound quality due to errors can be reduced.

Next, one embodiment of the present invention will be described using the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention, and FIG. 2 is a diagram of an example of input/output and signals during processing.

First, an input signal, which is a sample value of the original waveform inputted from the input terminal 107, is sent to the input buffer 101.
is stored in An example of an input signal whose sampling period is Ts is shown in FIG. 2a. According to the control signal sent from the control circuit 106 via the control signal transmission line A114, the input signal is transferred from the input buffer 101 to the zero interpolator 102 through the input signal transmission line 1.
10.

In the zero interpolator 102, the control circuit 106
According to the control signal sent from the input buffer 101 via the control signal transmission line B115, M-1 sample values having a value of zero are inserted between the sample values of the input signal sent from the input buffer 101, and as a result, the sampling The signal whose period has been multiplied by 1/M is sent to the digital low-pass filter 103 through the zero interpolation signal transmission line 111. FIG. 2b shows an example of a signal in which a sample having a value of zero is inserted into the input signal illustrated in FIG. 2a. Although this example is for M=2, generally the influence of errors is reduced as the value of M is large.

The digital low-pass filter 103 filters the zero interpolator 10 according to a control signal sent from the control circuit 106 via the control signal transmission path C116.
The signal sent from 2 is filtered, and the resulting signal is sent to linear interpolator 104 via filter output transmission line 112. FIG. 2c shows an example of the output signal of the digital low-pass filter.

In accordance with the control signal sent from the control circuit 106 via the control signal transmission line D117, the linear interpolator 104 divides the sample value for each period T _s ' between the samples of the signal sent from the digital low-pass filter 103. is obtained by linear interpolation, and the resulting signal is sent to the output buffer 105 via the linear interpolation signal transmission line 113. The control signal sent from the control circuit 106 to the linear interpolator 104 consists of information instructing to receive the output signal of the digital low-pass filter 103 and information instructing the time at which linear interpolation is to be performed. The latter information is obtained by the control circuit 106 based on the adjustment rate α input from the adjustment rate input terminal 109 and the sampling period Ts of the system.
From this, T _{s ′ is calculated and generated as T s ′} =α·T _s . The linear interpolation is shown in Figure 2c, and the resulting signal is shown in Figure 2d.

The output buffer 105 stores the signal sent from the linear interpolator 104, and outputs the signal to the output terminal 108 at every sampling period Ts according to the control signal sent from the control circuit 106 via the control signal transmission path E118. . The signal output from the output terminal is shown in FIG. 2e.

Note that although linear interpolation is performed only for the portion corresponding to the sampling period T _s ' in the embodiment, linear interpolation may be performed for all portions. However, in this case, the amount of processing will increase accordingly.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention;
FIGS. 2a to 2e are diagrams showing examples of input signals, output signals, and signals in the middle of processing. In the figure, 101 is an input buffer, 102 is a zero interpolator, 103 is a digital low-pass filter, 104 is a linear interpolator, 105 is an output buffer, 106 is a control circuit, 107 is an input terminal, 10
Reference numeral 8 indicates an output terminal, and reference numeral 109 indicates an adjustment rate input terminal.

Claims (1)

[Claims] 1 By interpolating between the sample values of the waveform sampled in the first cycle, a sample value is generated for each second cycle, and the pitch is changed by outputting the sample value in the first cycle. A pitch adjustment device of the type shown in FIG. a low-pass filter; and means for generating a new sample value by linear interpolation between the sample values before and after the time in each second cycle among the output sample values of the digital low-pass filter, and outputting the new sample value in the first cycle. A pitch adjustment device comprising: