JPH11126093A - Voice input adjusting method and voice input system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the recognition rate of a voice input and to facilitate the input work. SOLUTION: An A/D converting device 2 multiplies an input signal from a microphone 1 by gain to make an A/D conversion and a voice recognition device 3 pattern-matches with a word of a syllable dictionary while dividing the input signal into prescribed units (syllable) when the input signal is voice, to output a suited word code to a CPU 4. At this time, a level measurement part 32 measures the sound pressure level of the prescribed unit to output it to the CPU 4. A level decision part 42 in the CPU 4 decides whether or not the measured level lies, within a prescribed range obtainable a high recognition rate, and when out of the range, the level decision part 42 outputs an input gain control signal to a gain adjustment part 22 to control so that the input voice lies within the prescribed range.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice input system, and more particularly to a method for adjusting a voice input gain for voice recognition.

[0002]

2. Description of the Related Art A voice data input system which inputs measurement data, control commands, and the like by voice, converts the data into coded data by a voice recognition device, and inputs the coded data to a computer device has begun to spread. Voice input from word processors and map guidance systems for drivers has also entered the stage of practical use.

In a conventional voice input system, the gain (volume) of a voice input unit is adjusted to an appropriate range in accordance with the sound pressure level of a speaker's voice before formal voice input.

[0004]

In the conventional voice input adjustment method, the input gain after adjustment is fixed. For this reason,
When the speaker changes the volume (strength) of the voice based on the surrounding situation, matching with the voice dictionary may fail and recognition may not be possible. It was bad.

An object of the present invention is to provide a speech input adjustment method having a high recognition rate irrespective of the magnitude (strong or weak) of a speaker's input speech, and a user-friendly speech input system using the same.

[0006]

According to the present invention, an input signal from a voice input unit is A / D converted, and a voice dictionary storing code data corresponding to a voice pattern in units of words or the like is used. In a voice input method of performing pattern matching and converting an input signal into a suitable code, when the input signal includes a pause for a certain period of time or more, the input signal is determined to be voice and its sound pressure level is obtained. It is characterized in that it is determined whether the sound pressure is within a predetermined sound pressure range with a good recognition rate, and when the sound pressure is out of the range, the input gain is adjusted so as to be within the range.

In the voice input system for realizing the present invention, the voice input unit of the A / D converter has an automatic adjustment function of a voice input gain, the voice recognition device has a function of measuring a sound pressure level of an input voice signal, and the CPU. Each of which has a level range determination function, and outputs an input gain adjustment signal from the CPU to the automatic adjustment function when the sound pressure level of the input audio signal is out of a preset sound pressure range. Features.

According to the present invention, a stable and high recognition rate can be obtained even when the speaker changes and the volume of the voice is different, or when the voice is changed and input depending on the surrounding conditions. The work of re-input due to the failure of can be greatly reduced.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below in detail with reference to the drawings. FIG. 1 shows functional blocks of a voice input system according to one embodiment. The input signal from the microphone 1 is a temporal change in sound pressure. The A / D converter 2 multiplies the gain, and then converts the analog signal into a digital signal by the A / D conversion function 21, and the voice recognition device 3
Output to

The recognition processing unit 31 of the speech recognition device 3 divides the input signal into a syllable or a word based on a characteristic amount such as a segment rhythm or a pause while storing the input signal in a memory in a time series. Pattern matching is performed with a voice dictionary in units of words (for example, “Denatsu (voltage)”) combining “i”, “u”,..., And the recognized word code is output to the CPU 4. .

At this time, if there is no word corresponding to the voice input, a code representing no corresponding word is output. Also, when the input signal is a continuous sound, etc., for a certain period of time (for example, 0.4 seconds)
When the pause section is not included, the recognition process is stopped as if it were noise. The CPU 4 calculates the input word code according to the application program 41 or stores it in a memory. In addition, there are a plurality of well-known methods of voice recognition other than the above, and there is no particular limitation in the configuration of the present invention.

Further, in the voice input system according to the present embodiment, the gain adjustment section 22 is provided in the A / D converter 2, the level measurement section 32 is provided in the voice recognition apparatus 3, and the level determination section 42 is provided in the CPU 4.
And the automatic adjustment of the voice input gain is performed as follows.

FIG. 2 is a flow chart of a method for adjusting a voice input gain according to one embodiment. First, the recognition processing unit 31 determines whether or not the input signal is a voice (s101). If the input signal is a voice, the sound pressure level (average value) of the division unit by the pause unit is measured (s102).

Next, it is determined whether the measurement level is within a predetermined range set in advance, that is, a range suitable for recognition processing (s103). As a result of the determination, if the measurement level is higher than the predetermined range, a gain-down command corresponding to the deviation is output to the gain adjustment unit 22 (s104), and if the measurement level is smaller than the predetermined range, a gain-up command is output. (S105).
Also, if it is within the predetermined range, no gain adjustment command is output.

FIG. 3 shows a conceptual diagram of the operation of adjusting the voice input according to the present embodiment. When the level of the input speech 301 is high, the gain of the A / D converter 2 is reduced, and the subsequent input speech 302 is controlled so as to enter a high recognition rate range. When the level of the input voice 303 is low, the gain is increased, and the subsequent input voice 304 is controlled so as to be in the high recognition rate range.

As described above, the present embodiment automatically adjusts the gain in real time according to the actual level of the input voice so as to enter the level range having a high recognition rate. Can be improved, and re-input can be greatly reduced.

[0017]

According to the present invention, since the input gain is optimized in real time according to the input voice, the recognition rate of the input voice is improved, and the work of voice input is simplified.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a voice input system according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a voice input adjustment method according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing the effect of the adjustment movement of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Microphone, 2 ... A / D converter, 22 ... Gain adjustment part, 3 ... Voice recognition device, 31 ... Recognition processing part, 32 ... Level measurement part, 4 ... CPU, 41 ... Application program, 42 ... Level determination part .

Claims (2)

[Claims] An input signal from a voice input unit is A / D converted, and pattern matching is performed using a voice dictionary storing code data corresponding to a voice pattern in units of words or the like. In the voice input method of converting into a suitable code, when the input signal includes a pause for a predetermined time or more, the input signal is determined to be voice and its sound pressure level is obtained, and the sound pressure level is a predetermined sound pressure having a high recognition rate. A voice input adjustment method comprising: judging whether a value is within a range, and adjusting the input gain so as to be within the range when the value is outside the range. 2. A voice input microphone, an A / D converter for converting an input voice signal into a digital signal, and voice recognition for performing a recognition process on the converted input voice signal using a voice dictionary and outputting a suitable code. An audio input system comprising: a device; and a CPU for performing predetermined application processing by inputting the code, wherein an automatic adjustment function of an audio input gain is provided in a voice input unit of the A / D converter; The CPU has a function of measuring the sound pressure level of the input audio signal and the CPU has a level range determination function. When the sound pressure level of the input audio signal is out of the preset sound pressure range, the CPU automatically adjusts the function. A voice input system configured to output an input gain adjustment signal to the voice input system.