JPS59137999A - Voice recognition equipment - 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] Industrial applications The present invention relates to speech recognition technology.

Conventional configurations and their problems In speech recognition methods, a common method is to compare input speech with a standard pattern and output all patterns with the highest degree of similarity. In this case, it is typical to use a word or a phoneme as a standard pattern unit. Since the following explanation will be based on speech recognition for all speakers, a recognition method using phonemes as a unit is effective, and the following explanation will be based on phoneme recognition. However, the content of the present invention is not limited to the method of recognizing phonemes, and the same method can be used even when words or sentences are used as units.

FIG. 1 shows a complete diagram of the part of the conventional device that performs the phoneme subfunction. 1 is an analysis section, 2 is a phoneme recognition comparison section,
3 is a phoneme standard pattern storage section.

The storage unit 3 contains as many standard features and parameters corresponding to each phoneme as there are phonemes. The comparison unit 2 compares each standard pattern with the input feature parameter analyzed by the analysis unit 1τ, and the symbol or number of the phoneme in the standard pattern that has the highest degree of similarity to the input feature/parameter is used as the phoneme recognition result. Output.

Creating standard nokuturns in a system that targets all speakers requires data from many people.

Therefore, since it is not possible to record the data of many people under each usage environment, it is absolutely necessary to create the data using data recorded in a certain environment (for example, in a soundproof room). For this reason, the matching between the input speech in a noisy environment (in a usage environment) and the noise-free standard pattern was not successful, leading to a decrease in the recognition rate. 5g2 figure S/
The relationship between N ratio (signal-to-noise ratio) and phoneme 4 recognition rate is shown. Looking at this, the recognition rate in the absence of noise is 86.2%, but as the S/N ratio decreases, the recognition rate also decreases, and SZN
When the ratio is 6dB (decibel), it is 72.7% and 12.6%.
Conventional methods that do not take into account the effects of noise, which can also be reduced, are clearly problematic.

Purpose of the Invention The present invention was made in order to solve the above-mentioned conventional problems, and it creates a standard pattern taking noise into consideration.
The aim is to reduce misrecognition by using all of them to perform phoneme recognition.

Structure of the Invention In order to achieve this object, the present invention prepares a plurality of standard patterns corresponding to the types of noise, and selects the optimum standard pattern according to the environmental noise.

Is the agS diagram an implementation of the present invention? ll, 1 is an analysis section, 2 is a phoneme recognition comparison section, and these correspond to FIG. 3 is the noise level detection section, 4 is the standard pattern selection &
B, 6 is a standard Bakun storage section.

The standard pattern storage section 6, unlike the storage section 3 of FIG. 1, stores a plurality of standard patterns corresponding to the noise level. Reference numeral 6 is a switch for switching modes, and when it is turned to the a911 side, it is the recognition mode, and when it is turned to the b side, it is the environment learning mode.

Before using the voice recognition device (in a state where no voice is being input), the switch is switched to the eib side, and first only the noise in the usage environment is detected by the noise level detection section of the noise level detection section 3. 4 the standard butter 7 storage section 6 depending on the detected noise level.
Select the corresponding phoneme standard pattern from among. This standard pattern will be used for subsequent recognition. When using the voice recognition device, switch switch 6 to side a,
As in the case shown in the figure, the input speech is converted into characteristic parameters in the analysis section 1, and compared with a standard pattern in the comparison gl12 to recognize phonemes.

Next, we will explain how to create a standard pattern that takes all noise into account. Reduce the nature of noise in the usage environment1.
If a standard pattern was created by modifying the method, it would make a significant contribution to improving the recognition rate, but this method is not realistic because there are various types of environmental noise. Therefore, here we will create a standard pattern by modeling the characteristics of the noise.

It is known that the IM wave characteristic of environmental noise is the HOTH spectrum characteristic shown in the 4th N in general d1. A model noise having HOTH spectrum characteristics is created as a representative noise. This model noise and audio data recorded in a soundproof room are mixed so that the S/N ratio is a constant value to create audio data containing noise. Next, using this data, a phoneme standard pattern is created in the same manner as before. Such a procedure is performed for several S/N ratios, and a plurality of standard patterns are created and stored in the standard pattern storage section 6. In this example, the S/N ratio is 6d.
B to 36 dB'ii, standard patterns are created every 5 dB, and a total of 8 types of standard patterns are created and stored in the standard pattern storage iBs. The standard pattern created in this way is general-purpose, and once created, there is no need to change it.

The noise level detection unit 3 calculates the sum of squares of the noise signals input from the microphone. This is the part that calculates the power.

The standard pattern selection unit 4 converts the noise power into an S/N ratio, and selects the standard pattern closest to the corresponding S/N ratio from the standard pattern storage unit 6. Conversion between S/N ratio and noise power is performed using the following equation.

S/N ratio = 1010g (sound power) - 1010q (noise power) Here, the sound power input to the microphone (first term on the right side) can be considered to be approximately constant, so use the above equation.

It is possible to correlate S/N ratio and noise power.

Effects of the Invention The effects of the present invention will be evaluated based on the phoneme recognition rate. The phoneme recognition rate used here is the number of correctly recognized frames (1
A frame is defined as the ratio of 10m5ec length audio data) to the total number of frames.

FIG. 6 shows, as an example, five vowels and nasal sounds (/m/, /n/) in a noisy environment corresponding to an S/N ratio of 26 dB.

This figure shows the evaluation results for the sound.

The solid line 7 is the result when the standard pattern according to the present invention is used, and the broken line 8 is the result when the conventional standard pattern is used. The average recognition rate improved by 3.6%, and nasal sounds improved by 26%.

Therefore, it can be said that the effects of the present invention are significant.

Figure 6 shows the evaluation of phonemes in 212 words uttered by 10 men.
The amount of data is approximately 0 frames, and the result is sufficiently reliable.

As described above, the present invention is a relatively simple and general method, and is highly effective in improving the phoneme recognition rate.

[Brief explanation of drawings]

Figure 1 is a block diagram of a conventional method for phoneme recognition using standard pattern matching, Figure 2 is a diagram showing the relationship between S/N ratio and phoneme recognition rate, and Figure 3 is a block diagram of a conventional method for phoneme recognition using standard pattern matching. FIG. 4 is a block diagram of a speech recognition device, and FIG. 4 is a diagram showing a frequency spectrum of noise. FIG. 6 is a diagram showing a recognition rate by the device of the present invention in comparison with that by a conventional device. 1... Analysis section, 2... Phoneme recognition comparison section, 3... Noise level detection section, 4... Standard pattern selection section. 6...Standard pattern storage section, 6...Mode changeover switch. Name of agent: Patent attorney Toshio Nakao and 1 other person No. 1
Figure 2 Figure 41! ! 1 @ Mid-Tsuzuri (KH2) Fig. 5 /6'/ /e/ /(11/if /ltl
/N/j′

Claims (1)

[Claims] It has a means to compare all the sound and standard patterns, prepares multiple standard patterns according to the type of noise, learns the environmental noise before using the device, and selects one of the above standard patterns according to the learning results. A speech recognition device characterized in that one type of corresponding standard pattern is selected and this standard pattern is conveniently used to recognize speech uttered in a noisy environment.