JPH03274599A - On-vehicle speech recognition device - Google Patents

Abstract

PURPOSE:To obtain a high stable recognition rate by varying the gain of a variable gain microphone amplifier according to the distance between a microphone and the mouse of a speaker which is estimated from the position of a seat detected by a seat position detecting means. CONSTITUTION:The voice signal from the microphone 1 is amplified by the variable gain microphone amplifier 10. A sheet position detector 12 detects the position of the seat 4 and the value of the best gain is found from the detection output by using an optimum gain conversion table 13, the gain of the variable gain microphone amplifier 10 is varied, and the influence of a loudness difference with the distance 1 between the mouse 2a of the speaker 2 and microphone 1. This voice output is inputted to an automatic gain control circuit 11 and a loudness difference due to the loudness of the speaker's voice is removed to obtain a constant output value. Consequently, the high stable recognition rate can be obtained.

Description

[Detailed description of the invention] 1 string and a string The present invention relates to a vehicle-mounted voice recognition device.

When the volume of the voice input from the microphone connected to the voice recognition device changes, the recognition rate generally deteriorates. This is because the input cover pattern changes as the volume changes. Therefore, when creating a pattern, in order to keep the audio level constant in front of the feature extraction section, AGC
(Automatic gain control circuit)
6880) is inside the shell. As a result, even if there is a slight change in volume, changes in the pattern can be reduced. However, outside the range in which AGC acts (normally, the control range of AGC is several tens of decibels), the recognition rate also deteriorates because the volume changes.

By the way, since the microphone used for voice recognition is generally a headset type, the distance from the microphone to the mouth is relatively stable. In such a case, only the change in volume due to individuality is considered, and in many cases, the change in volume falls within the range of action of AGC.

However, in the case of in-vehicle voice recognition, the microphone is of a hands-free type, as shown in FIG. 2, and the distance between the microphone and the mouth becomes large.

That is, in FIG. 2, 1 is a microphone, 2 is a speaker (driver), 3 is a steering wheel of the car, and 4 is a seat where the speaker sits. For example, when microphone 1 is attached to the dash board of the driver's seat, When the short female driver 2 takes the driving position, the seat 4 is moved largely toward the microphone 1 side, and the distance between the microphone 1 and the driver's mouth 2a becomes shorter. On the other hand, in the case of a tall male driver 2, the seat 4 moves away from the microphone 1, so the distance between the microphone 1 and the mouth 2a becomes longer. - inside the shell, the distance from the microphone 1 is 2
Since the sound pressure is inversely proportional to the power of the sound pressure, for example, if the distance between microphone 1 and mouth 2a differs by two times between a woman and a man, the volume of the sound entering microphone 1 will differ by four times. Voice recognition is
Usually, there is a volume at which the recognition rate is best, and as the volume changes, the recognition rate gradually deteriorates, and furthermore, when the AGC control range is exceeded, the recognition rate deteriorates significantly. If the distance from the microphone 1 changes and the volume changes, it becomes difficult to obtain the volume that will give the best recognition rate. Also,
If this is combined with changes in volume due to individual characteristics, the volume tends to change beyond the control range of the AGC, making it difficult to obtain a stable recognition rate.

The present invention was made in view of the above-mentioned circumstances, and
In particular, the distance between the microphone and the speaker's mouth is estimated by detecting the position of the seat that the speaker has moved to take up his or her driving position, and the gain of the microphone amplifier is varied based on this distance. To provide an in-vehicle speech recognition device that can operate so that the output of a microphone amplifier remains constant even when the seat position is changed, and can stably maintain a high recognition rate even when used by speakers of two different physiques. It was done for that purpose.

In order to achieve the above object, the present invention provides (1) a microphone, a variable gain microphone amplifier to which an electrical signal from the microphone is input, and an AGC to which an output signal from the microphone amplifier is input. (automatic gain control circuit), etc., and seat position detection means for detecting the seat position where the speaker is sitting, and the speaker is estimated based on the seat position detected by the seat position detection means. The gain of the variable gain microphone amplifier is changed according to the distance between the mouth of the microphone and the microphone, and further, the gain of the variable gain microphone amplifier is changed according to the distance between the mouth of the microphone and the microphone.
), (2) the seat position is largely represented at several locations, and the gain of the microphone amplifier is represented by a certain gain in the vicinity thereof; or (3) a gain whose acoustic characteristics are opposite to each other. In other words, the microphone amplifier has a gain proportional to the square of the distance between the microphone and the speaker's mouth, or (4) if the distance between the microphone and the speaker's mouth is small, the The present invention is characterized in that the gain of the microphone amplifier is greater than the gain proportional to the square of the distance between the microphone and the speaker's mouth. Hereinafter, the present invention will be explained based on examples.

Figure 2 is a diagram showing the relationship between a speaker and a microphone when performing in-vehicle voice recognition.
When the driver (driver) 2 adopts an arbitrary driving position, the distance between the microphone 1 and the speaker's mouth 2a changes because the microphone 1 is at a fixed position.

Even for the same speaker, if the distance increases, the output of the microphone ↓ will decrease, and if it is less, it will increase. Therefore, if the distance is known, by changing the gain of the microphone amplifier 1 in advance, a constant microphone amplifier output can be obtained regardless of the driving position of the speaker. If the distance HL is small, reduce the gain of the microphone amplifier 6
On the other hand, if the distance is large, increase the gain of the microphone amplifier. The simplest way to determine the distance is to detect the position of the seat used by the speaker. In the case of a car, the seat position is the most important factor when determining the driving position. For people with a short distance, the seat 4 is always closer to the dash board side, and on the other hand, for people with a longer distance, the seat 4 is always closer to the rear seat side. Therefore, the distance can be determined from the seat position, and if the gain of the microphone amplifier is adjusted using this distance, a stable recognition rate can be obtained.

FIG. 1 is a configuration diagram for explaining one embodiment of the present invention, in which 10 is a variable gain microphone amplifier, 11 is an AGC
(automatic gain control circuit), 12 is a seat position detector, 13
is the optimal gain conversion table section.

14 is a gain control information output section, 15 is a feature extraction section, 16 is a speech section detection section, 17 is a dictionary creation section, 18 is a dictionary, 19 is a recognition section, 2o is a result display section, and the audio signal from the microphone 1 is The signal is amplified by the variable gain microphone amplifier 10, and the influence of the volume difference due to the distance between the speaker's mouth and the microphone is removed as described below. This audio output is input to the AGC cycle i% 11, the volume difference due to the loudness of the speaker's voice is removed, the output is set to a constant value, and the output is input to the feature extraction unit 15 and the voice section detection unit 16. . The speech pattern obtained here is transferred to the dictionary creation unit 17 and used as data for creating a dictionary, or is input to the recognition unit 19 that performs recognition by determining the degree of similarity with the dictionary 18. The word is output to the result display section 20 as a correct word.

Next, a method for determining the gain of the microphone amplifier 10 will be described. As described above, the distance ML from the microphone 1 to the speaker's mouth 2a is determined by the position of the seat 4. Therefore, the position of the seat 4 is detected by the seat position detector 12, and from the detection output, an optimal gain conversion table 13
is used to determine the optimum gain value, and the gain of the microphone amplifier 10 is controlled using this information. Hereinafter, the detection of the seat position and the control of the gain will be described using examples.

Figure 3 shows an example in which the seat positions are roughly divided into 3 to 4 groups and the gain of the microphone amplifier 0 is changed roughly according to the seat position. For example, as shown in the figure, a switch is attached to the seat, and these switches change the seat to A, B, or C.
The position of the microphone amplifier 10 is detected and the gain of the microphone amplifier 10 is changed to a, b, or c depending on the position. In addition, in Fig. 3, the seat positions are divided into three stages, A, B, and C, and the corresponding positions (
Gains (a, b, c) are predetermined for A, B, C) to form an optimal gain conversion table section, and the position of the switch functions as a gain control information output section 14. FIG. 4 is a diagram showing the relationship between the distance between the microphone 1 and the speaker's mouth 2a and the gain G of the microphone amplifier 10 in this case. When the distance between the two is small, reduce the gain,
When the distance is large, increase the gain and use microphone amplifier 1.
I try to get a nearly constant output from 0.

Figure 5 shows the position information (detection voltage) of seat 4 and converting it to a digital value.The position of seat 4 is detected as a voltage value using a variable resistor, etc., and the detected voltage is A/D converted. A/D conversion is performed using a device to create a digital value. ), the physical volume error is corrected using R○M13 for this digital value, and in this embodiment, the distance between the microphone 1 and the speaker's mouth 2a and the gain G of the microphone amplifier are As shown by curve A in FIG.
=kL", and the gain of the microphone amplifier 10 is determined using the analog information obtained by D/A converting this ROM value with a D/A converter. In other words, in this embodiment, the detection voltage section and the A/D conversion section are
The seat position detection means 12 shown in FIG.
M forms the optimal gain conversion table section 13, and the D/A conversion section forms the gain control information output section 14.

In general, people whose seats are forward are often short people (with women inside), and women's voices are lower in volume than men. Therefore, when the seat is in front, if the gain is lowered just because the distance 1iL between the microphone and the speaker's mouth is close, the voice uttered will be low and the output of the microphone amplifier will be smaller than expected. Therefore, for the values written in the table of ROM 13 of Example 2,
As shown by curve B in FIG. 7, the gain G is made slightly larger when the distance is small.

In this way, the volume can be corrected correctly even for women.

As is clear from the above explanation, according to the present invention, in an in-vehicle voice recognition device, the volume can be adjusted by the microphone amplifier output even if the speaker takes a different driving position depending on the height etc. Since the processing is performed so that the pattern is constant, there is little variation in the pattern and a stable and high recognition rate can be obtained.

In addition, when a woman uses the device, it operates to compensate for the fact that the woman's voice is low, so a stable and high recognition rate can be obtained.

[Brief explanation of drawings]

FIG. 1 is a block diagram for explaining one embodiment of the present invention, FIG. 2 is a diagram for explaining detailed explanation of the present invention, and FIG.
4 and 4 are diagrams for explaining an example of the gain adjustment of the microphone amplifier, and FIGS. 5 to 7 are diagrams for explaining other examples of the gain adjustment of the microphone amplifier. Engineering...Microphone, 2...Driver, 3...Steering wheel, 4...Seat, 10...Variable gain microphone amplifier. 11...AGC, 12...Seat position detector, 13
... Optimum gain conversion table section, 14... Gain control information output section, Engineering 5... Feature extraction section, 16...
- Phonetic section detection section, 17... Dictionary creation section, 18. Dictionary,
Engineering 9: Recognition unit, 20: Result display unit. Figure 1 Figure 3 803-

Claims (1)

[Claims] 1. A voice recognition device including a microphone, a variable gain microphone amplifier into which an electrical signal from the microphone is input, an AGC (automatic gain control circuit), etc. into which an output signal from the microamp is input, and a voice recognition device in which a speaker is seat position detection means for detecting the position of a seat on which the person is sitting; and the gain is variable in accordance with the distance between the speaker's mouth and the microphone, which is estimated based on the seat position detected by the seat position detection means. An in-vehicle voice recognition device that is characterized by changing the gain of the microphone amplifier.