JPH02101500A - Voice recognizing device - Google Patents

Abstract

PURPOSE:To efficiently instruct a voice producing level to a talker so that the voice of the talker can be recognized at a high recognition rate by comparing an input level with two or more threshold levels and informing the compared results to the talker in the form of short alarming sounds. CONSTITUTION:This voice recognizing means is provided with a means 12 which compares an input level with two or more threshold levels and another means 14 which generates to or more kinds of short alarming sounds and, when a voice input is made, the input is compared with the threshold levels. Then, when the input does not reach the lower threshold level, an alarming sound indicating that the inputted voices are low is produced and, when the input exceeds the higher threshold, another sound indicating that the inputted voices are loud is produced. Therefore, even when a talker cannot see an indicator, etc., due to sound dial, etc., of an automobile telephone set while he drives a car, he can recognize whether or not his voice level is appropriate by the alarming sounds. Thus a high recognition rate can be achieved.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice recognition device, and is applicable to various devices such as voice dialing of a car phone, for example.

Figure 2 of the Seven Views of Jubei shows the conventional voice recognition method
In the figure, 1 is a microphone, 2 is an input level Kil+ constant/judgment section, and 3 is a diagram for explaining an example of Publication No. 0).
is a voice recognition unit, 4 is a first lamp, 5 is a second lamp, and the voice input from microphone 1 is input level 1.
11! The level is measured by the I-determination/judgment section 2, and the speech recognition section 3
3 is recognized. 2. To judge the power level, the input level 1fl11 constant/judgment unit 2 has two level thresholds, and it makes a judgment on the magnitude of these two thresholds, and based on the judgment result, the two lamps 4
, 5.

That is, when the input level is less than the lower threshold, the lamp is not turned on, when it is between the two thresholds, only the first lamp 4 is turned on, and when it is greater than the higher threshold, the lamp 4 is turned on. 4.5 lamps are lit at the same time. This input level determination and lamp control continues in the same way not only when <y2yl is performed, but also when registering a standard pattern for recognition.

Here, the two thresholds are set such that the higher threshold is the maximum level of the A/D converter of the speech recognition unit 3, and the lower threshold is approximately 25 dB lower than the higher threshold.

Although such input level determination and lamp blinking are not directly related to the speech recognition algorithm, they encourage optimization of the speech level and ensure a high recognition rate. In other words, if the vocalization level is too high, the sound may be distorted in the input system or A/
Failures such as saturation of the D converter may occur, and if it is too low, the S/N with the surrounding noise will become sluggish, and the recognition rate will decrease in any case, so the speaker should not look at the lamp while speaking. It was necessary to vocalize while detecting whether or not the level was at an appropriate level.

However, in the above conventional speech recognition method.

When voice recognition is applied to telephone voice dialing, etc., the person speaking needs to look at the lamp, and if the person is unable to do so, the person speaking cannot confirm the appropriate level and may not be able to get good recognition. It was not possible to obtain the jil rate.

L-1 The present invention was made in order to solve the above-mentioned problems, and was made for the purpose of providing a speech recognition device that can ensure confirmation of the speaking level of a single speaker and obtain a high recognition rate. It is something.

31-Ikka In order to achieve the above object, the present invention sets the input level to 2.
and a means for generating two or more types of short warning sounds. When the input audio is below a threshold, a warning sound is generated to indicate that the input audio is low, and when the input is above a higher threshold, a warning sound is generated to indicate that the input audio is loud. It is. Embodiments of the present invention will be explained below.

Therefore, the present invention changes the input level to di! 2 steps with I fixed
It has means for comparing with a threshold of 1 TE or more, and means for generating a short warning sound that can be distinguished by two or more types of humans, and when there is a voice input, it compares the input with the threshold, When the input is below the lower threshold, a warning sound is generated to indicate that the input audio is low, and when the input is above the high threshold, a warning sound is generated to indicate that the input audio is loud. It is something. Now, think about the meaning of using a warning sound. Usually, the output of the recognition result of the speech recognition device n uses text or speech information such as a display or a voice response to indicate candidate categories. Voice can be applied as the means in question here, and while it is effective to notify the vocalization status using rt voice when using the voice recognition device for the first time, when using it constantly, voice The instructions are:
For example, since it takes a certain amount of time to say something like "Please speak loudly" or "Your voice is too loud," this sound becomes annoying. Therefore, by generating a warning sound indicating the vocalization level as in the present invention.

The speaker can be notified in a short time.

Therefore, according to the present invention, it is possible to know whether the level is appropriate or not by the warning sound even when the person making the call is driving and cannot see the display, etc. by voice dialing of a car phone, etc. It has the effect of achieving a high recognition rate.

FIG. 1 is a block diagram for explaining one embodiment of the present invention, in which 11 is a microphone, 12 is an input level measurement/judgment section, 13 is a voice recognition section, 14 is a warning sound generation section, 1
Reference numeral 5 denotes a speaker, and the level of the voice input from the microphone 11 is measured by an input level measuring/judgment section 12, and recognized by a voice recognition section 13. The determination of this input level is
The input level measurement/judgment unit 12 has two or more level thresholds, and determines the magnitude of the two or more thresholds, and sends a control signal to the warning sound generation unit 14 based on the determination result. Based on the command, the warning sound generator 14 generates a sound that can be recognized by one person (for example, approximately 200
A short tone (of the order of ms) is output to the speaker 15 with a different tone.

In addition, when the input level is lower than the lower threshold, it uses the human tendency to become louder when hearing a soft voice or sound, such as on a telephone, and to become softer when hearing a loud voice. The vocalization can also be controlled by decreasing the output level from speaker 15 and increasing the output level from speaker 15 when it is greater than a higher threshold.

It goes without saying that the input level determination and warning sound control are performed not only when performing recognition, but also when registering a standard pattern for recognition.

As is clear from the explanation in ``Effects'' 2, according to the present invention, the input level is compared with two or more thresholds, and the result is notified to the speaker as a short-term warning sound, thereby efficiently achieving the desired effect. It is possible to instruct the speaker about the state of the speech level, and a high recognition rate can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram for explaining an embodiment of a speech recognition apparatus according to the present invention, and FIG. 2 is a block diagram for explaining an example of a conventional speech recognition apparatus. DESCRIPTION OF SYMBOLS 11... Microphone, 12... Input level measurement/judgment part, 13... Voice recognition part, 14... Warning sound generation part. 15...Speaker.

Claims (1)

[Claims] 1. It has means for comparing the input level with two or more thresholds, and means for generating two or more types of short warning sounds, and when there is a voice input, the input is compared with the thresholds, When the input is below the lower threshold, a warning sound is generated to indicate that the input audio is low, and when the input is above the high threshold, a warning sound is generated to indicate that the input audio is loud. A speech recognition device characterized by: