SU619953A1 - Device for identification of speech signals - Google Patents

Description

The invention relates to the field of accus TIC0 and can be used to enter information into electronic computers and automated control systems.

There are known devices in which the extrusion of signs of a signal occurs on the basis of an analysis of its TlI spectrum.

As a rule, frequency analysis blocks, which are a system of band-pass or resonant filters, are used in the devices. However, analog filters are characterized by significant transients, frequency response instability, and adjustment complexity.

Devices are also known that use digital filtering to distinguish the frequency properties of a speech signal, ensuring the stability of characteristics and the simplicity of implementing filters on discrete logic elements.

Devices include microphone, filtered signal processing unit, memory unit, decision unit, device control unit. To implement

Digital filtering The speech signal is pre-sampled using an analog-digital converter, and then fed to a discrete filtering unit.

The disadvantages of these devices are the presence of such a complex block as an analog-to-digital converter, as well as the fact that the parameters of the speech signal allocated at the output of the discrete filtering unit are significantly dependent on the speaker's loudness. This complicates the process of comparing the current cHfHal parameters with the reference one and, therefore, reduces the reliability of recognition

The aim of the invention is to provide a portable device by using clippings.

This is achieved by the fact that the proposed device contains a clipping unit, the input of which is connected to the output of the microphone and the output to the discrete filtering unit and the control unit.

Claims (2)

In addition, the discrete filtering block consists of ttt parallel to the block of blocking the block of determining the coefficients of expansion according to the number of signs of the speech signal, the Walsh function generator, the output of which is connected to the second inputs of the block of definition. decomposition coefficients, and a control unit, one output of which is connected to the third inputs of the decomposition coefficient determination units, the other to the input of the Walsh function generator, and its input is connected to the device control unit, and the generator of Walsh functions whose output is connected to the second inputs decomposition coefficient determination units; and the third inputs of the latter are connected to the output of the control unit, to which the input of the Walsh generator is connected. The circuit of the described device is shown in FIG. 1. It contains microphone 1, clipping unit 2, discrete filtering unit 3, Walsh function generator, 4, control unit 5, decomposition coefficient determination part 6, filter signal processing blog 7, received memory block 8, solutions and control unit 10. The output of block 1 is connected to the input of block 2, the output of block 2 is connected to block 10 and blocks 6, the output of block 10 is connected to blocks 5, 7, 8, 9, the first output of block 5 is connected to the input of block 4, the corresponding outputs of block 4 , .. dinene with the input of the corresponding block 6. The second output of block 5 is connected to. The inputs of the blocks 6, the outputs of the blocks are connected to the input of block 7, the output of block 7 is connected to the first input of block 8, the output of block 8 is connected to the input of block 9, and the output of block 9 is connected to the second input of block 8. Speech recognition device The campaign works as follows. The speech signal from microphone 1 is fed to a clipping unit. The speech, which is the clipping of the signal. As is well known, speech after Klippirovanin retains its intelligibility, and the whole process is concentrated in the passage of the zip signal. Selected features do not depend on an intense signal. The signal after clipping is fed to block 3 of discrete filtering, which implements a fast Walsh-Fourier transform on the interval Tg (time base). On the Tg interval (FIG. 2), the original signal can be represented by a sequence of average values of the function on the subinterval, moreover, in accordance with the Kotelnikov Ro theorem, the speech signal CS can be chosen equal to 4 kHz. Consequently, to decompose a speech signal into Walsh orthogonal functions, it is necessary that in the system of functions used in the last lastots, the Walsh function with maximum frequency has a duration between two zero-crossing (t.). (2) The number of functions used to decompose is always KpaTiCo two and is equal to -r / m (h) where / U is the number of the Walsh function Soil corresponding to condition (2). Consequently, the value of the interval Tg must be selected from the condition of the relation. Let, for example, Tg 2 msec, atsts (7 kHz, then 125 microseconds. From relation (2) we choose t 125 microseconds, and substituting in the expression (4), we obtain therefore, with a time base of% ms and ..Anax signal decomposition should use the system of Walsh functions with parts from CaB (0, в) (8.6), where © is the normalized time. When a signal is received at the output of the control unit, the control unit 10 issues a control signal 5 control. The latter triggers the Walsh 4 function generator, which during the Tg interval forms The Walsh function system. For each block for determining the coefficients of decomposition 6, the corresponding Wash function is applied. At the end of the Tg interval, the resulting set of coefficients is supplied to the filtered signal processing unit 7, which reduces the description of the speech signal in the Thickness factor space. memory block 8 and is used by decision block 9 either as a reference (learning mode) or for comparison with a reference set (recognition mode). Thus, the proposed device makes it possible to increase the reliability of recognition by distinguishing more stable speech signal parameters independent of the speaker’s speech volume, and also to simplify the recognition device by eliminating the analog-to-digital conversion unit. Claim 1. A speech recognition device comprising a microphone, a unit with a device and a sequence in a connected discrete filtering unit, a discrete information processing unit, a memory unit and an nfoaiifriva unit. solutions parallely connected to the output of the unit by a device, which is, in order to simplify the device by using clippy speech, the device contains a highlight, which is connected to the output of the microphone and the output to the unit of the unit (module and control unit. 2. The device according to claim 1, of which there is a discrete block of 6 3 radio sets consisting of W parallel to the unit clipping blocks for determining and dividing the expansion coefficients by the number of signs of a speech signal, P1 Walsh functions, the output of which is connected to the second inputs of the artifact blocks and no decomposition coefficients, and the control unit, one output of which is connected to the third inputs of the decomposition coefficient determination units, the other to the Walsh function generator, and its input is connected to the device control unit and the Walsh function generator, the output of which is connected to the second inputs of the decomposition coefficient determination units, and the last three inputs are connected to the output of the control unit, to which the Wal generator input is connected. a. Sources of information taken into account in the examination: 1. USSR author's certificate No. 304615, cl. Q 10 L 1 / O2, 1972. 2. USSR author's certificate number 441585, cells / S b 1/02, 1974. fpu.f