JP2658649B2 - In-vehicle voice dialer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mobile telephone terminal, and more particularly to a voice dialer for dialing with a voice signal.

[0002]

2. Description of the Related Art In a car telephone, a hands-free telephone has been attracting attention from the viewpoint of safety during traveling. Furthermore, since the operation is hands-free, a dial number registered in advance by voice is used. Voice dialers that can make phone calls are drawing attention. This voice dialer is provided with a voice analysis unit and a voice recognition unit, and most of the voice analysis units perform frequency analysis, and are configured by band filters of a plurality of channels. And, as the voice recognition unit, for example, DP
The specific speaker recognition unit of the matching method is used, detects the voice section of the analyzed voice, calculates the distance to the standard pattern registered in advance, and checks the similarity, and the highest similarity is obtained. Recognition is performed as a standard pattern and dialing is performed.

[0003]

In such a voice dialer of a car telephone, it is necessary to register its own voice in advance. This registration is usually performed in a relatively quiet environment while the vehicle is stopped. However, since the actual use is performed in the noise environment of a running automobile, an error in comparing the two is large. In particular, the noise in a vehicle has a characteristic of 1 / f, and the influence of the noise in the low frequency region becomes remarkable at high speeds due to the noise generated with the speed of the vehicle during traveling. For this reason, there is a problem in that an error in voice section detection particularly in a low band channel becomes large, so that it is difficult to cut out voice and the recognition rate is deteriorated. An object of the present invention is to provide an in-vehicle voice dialer with improved voice section detection, particularly in a low-band channel, and an improved recognition rate during driving of a car.

[0004]

A vehicle-mounted voice dialer according to the present invention includes a voice section detection section for detecting a voice section of a voice signal from a hands-free microphone analyzed for each frequency channel by a frequency analysis section. A vehicle speed sensor is connected, and each frequency channel signal from the frequency analysis unit is weighted according to the vehicle speed signal from the vehicle speed sensor. In this case, the weighting coefficient in the low band channel is made smaller than that in the high band channel.

[0005]

According to the present invention, by weighting each frequency channel signal in accordance with the vehicle speed of the vehicle, the effect of noise caused by the vehicle speed is improved and the speech recognition rate is improved.

[0006]

Next, the present invention will be described with reference to the drawings. FIG. 2 is a block diagram of a mobile telephone device using the present invention. In the figure, 1 is a hands-free microphone, 2 is a voice dialer, 3 is a control unit for controlling a car telephone device, 4 is a radio unit, and 5 is a vehicle speed sensor. FIG. 1 is a block diagram showing the internal structure of the voice dialer 2, and an amplifying unit 6 for amplifying a signal from the hands-free microphone 1.
A which converts the amplified signal into a digital signal
A / D converter 7, a frequency analyzer 8 including N band filters for analyzing the converted signal into N channel frequencies, and a start and end of voice based on the analysis result in the frequency analyzer 8. And a voice recognition unit 10 that receives the voice pattern specified by the voice section detection unit 9 and compares it with a pre-registered standard pattern for recognition. Note that a vehicle speed signal from the vehicle speed sensor 5 is input to the voice section detection unit 9.

In this configuration, first, at the time of registration, the audio signal input from the hands-free microphone 1 is converted into a digital signal by the A / D converter 7 in the same manner as in the normal specific speaker type speech recognition. This signal is subjected to frequency analysis by N band filters provided in the frequency analysis unit 8, and the result of the frequency analysis is sent to the voice section detection unit 9. The voice section detection unit 9 calculates the sum of each channel based on the frequency-analyzed signal and calculates the power. As a result, when it is determined that the speech is at the start end, the pattern of each channel from that time to the end is registered in the registration area of the speech recognition unit 10 as a standard pattern.

Next, a case where speech recognition is actually performed will be described. As in the case of registration, the generated audio signal is input to the A / D converter 7, and the output is input to the frequency analysis unit 8. Here, the output of the frequency analysis unit 8 performs the frequency analysis of the generated voice almost faithfully in a quiet environment such as at the time of registration, but 1 / f noise is dominant during running, and the low range is raised. Characteristics. For this reason, the voice section detection unit 9 performs voice section detection from this frequency analysis signal in the same manner as at the time of registration. At this time, the power calculation is performed by reducing the weight of the low-frequency channel by the vehicle speed signal. For example, assume that the bandpass filter is configured with 1/5 octave and 22 channels. Here, low frequency 1CH (center frequency: 250H
_Z), 2CH (center frequency: 287H _Z), 3CH (center frequency: If the 330H _Z), performs weighting corresponding to the vehicle speed with respect to these low frequency channel.

In this case, an increase in noise in the low frequency channel with respect to the vehicle speed is predicted in advance from data collected by a hands-free microphone at a speed of 20 to 100 km / h, and based on this, weighting of each channel at each vehicle speed is performed. Set. P = _K1CH × _P1CH + _K2CH × _P2CH + _K3CH × _P3CH
+ Where K is a weighting coefficient of each channel, P _iCH
Is the power of each channel. This K is changed according to the vehicle speed signal from the vehicle speed sensor so that the coefficient in the low band channel decreases as the speed increases. Next, an average of the power from several samples before is taken, and this is set as P ^AU . When you try exceeds a certain threshold amount of change in P ^AU the end of speech, the influence of low-frequency noise at high speeds is reduced, the variation of the audio signal itself is enhanced. As a result, even during high-speed running, voice section detection can be performed while suppressing the influence of running noise, and an improvement in the recognition rate during running can be expected.

[0010]

As described above, the present invention weights the signal from the frequency analysis unit in accordance with the vehicle speed signal from the vehicle speed sensor and performs voice section detection, thereby making it possible to reduce the noise in the vehicle especially during high-speed driving. This has the effect of suppressing the influence and improving the recognition rate of the voice for performing the voice dialer.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the internal structure of a voice dialer according to the present invention.

FIG. 2 is a block diagram of a car telephone device to which the present invention is applied.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Hands-free microphone 2 Voice dialer 3 Control part 4 Radio part 5 Vehicle speed sensor 8 Frequency analysis part 9 Voice section detection part 10 Voice recognition part

Claims (2)

(57) [Claims] 1. An automobile telephone apparatus for performing voice recognition by comparing a voice signal input to a hands-free microphone with a voice signal registered in advance and performing dialing based on a result of the voice recognition. Registers a frequency analysis unit that analyzes signals for each frequency channel, a voice section detection unit that performs voice section detection from voice signals sent from this frequency analysis unit, and a voice signal in the section specified by the voice section detection unit. A voice recognition unit that performs voice recognition in comparison with the obtained voice signal, wherein the voice section detection unit is connected to a vehicle speed sensor of the vehicle, and receives a signal from the frequency analysis unit according to a vehicle speed signal from the vehicle speed sensor. An in-vehicle voice dialer characterized in that weighting is performed on a frequency channel signal. 2. The on-vehicle voice dialer according to claim 1, wherein a weighting coefficient in the low band channel is smaller than that in the high band channel.