JPS61209500A - Voice signal conversion circuit - Google Patents

Abstract

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

Description

[Detailed Description of the Invention] [Summary] This circuit converts an analog audio signal into a digital audio signal, and is provided with a function to timely and automatically correct fluctuations in the zero point accompanying the conversion.

[Industrial application field]

The present invention is equipped in a speech recognition device using a computer, and is an audio signal converter that samples an analog audio signal inputted by a microphone or the like at predetermined time intervals and converts it into a time-series digital audio signal. It is related to circuits.

For example, in voice processing such as voice recognition, voice synthesis, and speaker recognition, features are extracted from voice prior to processing.

Various features are used as the above-mentioned features, such as an autocorrelation function, a short-time cass vector, a linear predicted number of needles, and a fundamental frequency. When extracting these features from an audio signal, first, for example, they are extracted at the timing of gkllz. The data is sampled and each sample value is converted into, for example, a binary 8-digit digital value by ^/D conversion, thereby converting it into a time series of bit string data. Desired characteristics are then calculated by a dedicated processing device containing, for example, a microcomputer.

By the way, 2. For example, the recognition accuracy in voice recognition is obviously affected by the extraction accuracy of the features used, and in particular, analog circuits are used for the AD conversion, and the zero point is likely to fluctuate due to temperature. Therefore, it is necessary to provide a means to correct this.

[Conventional technology]

In conventional audio signal conversion circuits, the AD conversion section is equipped with a zero point correction device such as a variable resistor.

Zero point correction was performed manually.

[Problem that the invention seeks to solve]

The audio signal conversion circuit having the above configuration has a problem in that it requires a great deal of effort to perform correction accurately.

[Means for solving problems]

The present invention includes a correction section that performs correction by addition and subtraction on a digital audio signal converted by an AD conversion section, and a correction section that integrates each sample value while the correction section continuously outputs a predetermined number of sample values that are less than or equal to a predetermined value. The above-mentioned problem is solved by including an integrating section that calculates the integrated value, and an average value calculating section that calculates the average value of the integrated integrated values.

[Effect]

In other words, the output of the correction section is monitored, the range in which it is below a predetermined value is considered to be a silent range, and when a silent section continues for a predetermined period of time, the average value of the output in that section is calculated, and this is sent to the correction section. By giving feedback,
This automatically corrects the zero point.

〔Example〕

Hereinafter, the gist of the present invention will be specifically explained using examples.

The figure is a block diagram showing the configuration of an embodiment of the present invention.

1 is the input analog audio signal, for example 1257
1! An AD conversion unit that samples every l and converts each sample value into an 8-digit binary digital value, thereby converting it into a digital audio signal expressed as a time series of bit string data.

Reference numeral 2 denotes a correction unit that performs correction by addition and subtraction on each sample value of the digital audio signal converted by the AD conversion unit 1, using an average value A obtained in an average value calculation unit 6, which will be described later.

3, the absolute value of the output of the correction unit 2 is experimentally determined as a threshold value for distinguishing between speech and silence, taking into account the level of noise in the analog audio signal input to the AD conversion unit 1. The comparison unit outputs V-1 when the dark value Tx is exceeded, and outputs V=0 when the dark value Tx is not exceeded.

4 is a binary N-digit (e.g., 8 binary digits) counter that clears the value when the output V of the comparator 3 is 1, and counts the sampling clock when the output V is 0;

5 is an integration unit that calculates the integrated value of each sample value while the correction unit 2 continuously outputs a predetermined number of sample values that are less than or equal to a predetermined value;
When the output V of the comparator 3 is 1, the integrated value is cleared and ■ becomes 0.
When , the output of the correction section 2 is integrated, and the count value of the counting section 4 is 2.
When the value reaches the Nth power (for example, 2 to the 8th power, that is, 256), the output signal E outputs the integrated value S.

Reference numeral 6 denotes an average value calculating unit that calculates the average value of the integrated values integrated in the integrating unit 5, and by shifting the integrated value S given from the integrating unit 5 by N (for example, 8) bits to the lower digit side. An average value A is obtained.

That is, when 2ON (for example, 2 to the 8th power, that is, 256) sample values that are less than or equal to the predetermined darkness value are successively output from the correction unit 2, the average value A is outputted from the average value calculation unit 6.
is fed back to the correction unit 2, and the value obtained by adding A to the current correction value X is set as a new correction value X, and the AD conversion unit 1
The output X of is corrected as follows.

For example, if the sampling period is 125 μs and the number of digits in the counter 4 is 8, the zero point is corrected every time a silent period continues for 32 minutes.

〔Effect of the invention〕

As described above, according to the present invention, the zero point drift in the AD converter is automatically corrected in a timely manner, and an accurate digital audio signal can be obtained without any effort.

[Brief explanation of the drawing]

The figure is a block diagram of one embodiment of the present invention. In fig. 1 is an AD conversion section, 2 is a data correction section. 3 is a comparison part, 4 is a counting part. 5 is an integration section, and 6 is an average value calculation section. A lot of old land! 10) Block 20

Claims (1)

[Claims] In an audio signal conversion circuit including an AD converter (1) that converts an input analog audio signal into a digital audio signal, each sample value of the converted digital audio signal is corrected by addition and subtraction. a correction unit (2) that performs the correction, and the correction unit (2)
an integrating unit (5) that integrates each sample value while continuously outputting a predetermined number of sample values that are less than or equal to a predetermined value; and an average value calculating unit (6) that calculates an average value of the integrated integrated values. An audio signal conversion circuit comprising: