JP2009181025A - On-vehicle speech recognition device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve a speech recognition rate without useless processing load and power comsumption, by stopping echo cancel processing by an echo canceler, when the speech recognition rate is hardly improved by echo canceling of the echo canceler. <P>SOLUTION: A control section 17 monitors a signal which is input to an echo cancel section 14, and controls the echo cancel processing by the echo cancel section 14 by determining whether or not, the speech recognition rate is improved by canceling an echo component. An echo suppression amount, a sound volume (an output level) and a noise level etc. may be monitored. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an in-vehicle speech recognition apparatus that includes an echo canceller that removes echo components contained in input speech and performs speech recognition.

With the progress of voice recognition processing technology, voice recognition devices have been applied to many fields, for example, in vehicles, they have been installed in navigation devices.
For this reason, for example, when performing destination setting, the navigation device recognizes the input voice by voice input of an address or the like instead of input from a touch screen or the like, so that the destination can be automatically set. It can also contribute to safe driving.

The above voice recognition is realized by picking up the user's input voice with a microphone, but in the vehicle, other music recorded on a CD (Compact Disc) or DVD (Digital Versatile Disc) is loaded into an audio device. In addition, since environmental noise such as running noise due to friction of air conditioners and tires may be picked up, the situation is severe in terms of recognition rate compared to normal usage scenes.
For this reason, an echo canceller is used in the speech recognition device, and echo cancellation processing is performed with this echo canceller to remove acoustic echo generated when sound emitted from other sound sources wraps around the microphone, Processing to reduce noise as much as possible is performed.

The echo cancellation processing described above suppresses the acoustic echo by estimating the characteristics of the transmission path of the sound in which the acoustic echo is generated and reproducing the sound transmission characteristics, and this acoustic echo suppression method has been conventionally used. Many patent applications have been filed.
For example, based on the vehicle state information collected by the state information input interface, the power SN ratio of the audio device output sound and the noise sound included in the microphone output signal, and the variation amount of the transfer function of the audio device output sound signal transmission system An in-vehicle audio processing apparatus that performs adaptive operation by controlling the characteristic of tap coefficient update operation in a FIR (Finite Impulse Response) filter is known (for example, see Patent Document 1).

JP 2007-19595 A

According to the technique disclosed in Patent Document 1 described above, since the adaptive operation filter for echo canceller performs a learning operation, even when the noise level is large or the transfer function fluctuates greatly, the calculation amount is not greatly increased. Adaptive operation can be performed with high accuracy.
However, the adaptive filter for the echo canceller does not necessarily work to improve the speech recognition rate, and from the viewpoint of the speech recognition rate, side effects due to the adaptive filter cannot be prevented. That is, for example, when the amount of echo cancellation is relatively large, or when other sound sources are output, the speech recognition rate is expected to be improved by the echo canceller, but when the amount of echo cancellation is extremely small, or other sound sources When there is no output, echo cancellation processing is executed even though the improvement of the speech recognition rate is not expected. In this case, useless processing load and power consumption of the speech recognition apparatus are involved.

  The present invention has been made to solve the above-described problems, and when the improvement of the speech recognition rate cannot be expected by the echo removal of the echo canceller, the echo cancellation processing by the echo canceller is stopped, thereby causing a wasteful processing load. Another object of the present invention is to provide a vehicle-mounted speech recognition device capable of improving the speech recognition rate without accompanying power consumption.

  In order to solve the above-described problem, an in-vehicle speech recognition device according to the present invention is an in-vehicle speech recognition device including an echo cancellation unit that performs an echo cancellation process for removing an echo component included in input speech, A sound source, an input speech collected via a microphone and a signal from the sound source as input, an echo canceling unit for removing an echo component contained in the signal, and speech recognition for recognizing the output of the echo canceling unit And a control unit that monitors a signal input to the echo cancellation unit, determines whether or not a speech recognition rate is improved by removing an echo component, and controls echo cancellation processing by the echo cancellation unit. It is provided.

  According to the in-vehicle speech recognition device of the present invention, when an improvement in the speech recognition rate cannot be expected due to the echo removal of the echo canceller (echo cancellation unit), it is a wasteful processing load by stopping the echo cancellation processing by the echo canceller. In addition, the speech recognition rate can be improved without power consumption.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing the internal configuration of a vehicle-mounted speech recognition apparatus according to an embodiment of the present invention.
As shown in FIG. 1, an in-vehicle speech recognition apparatus 1 according to an embodiment of the present invention includes a sound source 11, a sound output unit 12, a sound input unit 13, an echo cancellation unit 14, and a speech recognition unit 15. And a recognition result processing unit 16 and a control unit 17 having an echo cancellation amount determination unit 170.

The sound source 11 outputs output sound signals such as music, radio, and guidance sound to the sound output unit 12 and the echo cancellation unit 14. The sound output unit 12 is a speaker that outputs a signal output from the sound source 11 to the outside.
The voice input unit 13 acquires a voice collected via a microphone (not shown) as an input signal and outputs it to the echo cancellation unit 14. The echo cancellation unit 14 removes an echo of the sound output sound included in the sound signal output from the sound input unit 13 and outputs it to the sound recognition unit 15. Here, the echo cancellation unit 14 is set by the control unit 17 described later. It is assumed that there are an operation mode in which echo cancellation processing is enabled and a non-operation mode in which it is disabled.

  The voice recognition unit 15 performs voice recognition processing on the signal output from the echo cancellation unit 14 and delivers the recognition result to the recognition result processing unit 16. The speech recognition unit 15 also outputs status information indicating whether or not the speech recognition unit 15 is performing speech recognition processing. In order for the speech recognition unit 15 to perform speech recognition, an unillustrated acoustic model and a recognition dictionary are prepared as a database. The acoustic model includes acoustic characteristics of a small unit (phoneme) of human utterance, and the recognition dictionary includes Each word to be recognized is described. The voice recognition unit 15 calculates the acoustic characteristics by analyzing the voice signal delivered by the voice input unit 13 and taken in via the echo cancellation unit 14 by Fourier analysis or the like, and from among the words described in the word dictionary. A word whose acoustic characteristics are close to that of the input speech is searched for and delivered to the recognition result processing unit 16 as a speech recognition result.

The recognition result processing unit 16 is a function execution unit that displays a voice recognition result on a display unit (not shown) and sets a destination according to the voice recognition result output from the voice recognition unit 15.
A control unit 17 having an echo cancellation amount determination unit 170 (hereinafter simply referred to as control unit 17) detects an echo removal level, and determines whether or not to perform an echo cancellation process according to the detected removal level. The echo cancel unit 14 (operation mode / non-operation mode) is controlled. FIG. 2 is a flowchart showing the control procedure.

  Note that each of the echo cancel unit 14, the voice recognition unit 15, the recognition result processing unit 16, and the control unit 17 includes a microprocessor, and sequentially reads and executes programs recorded in the built-in memory. Assume that each of the functions described above is executed.

FIG. 2 is a flowchart showing the operation of the in-vehicle speech recognition apparatus according to Embodiment 1 of the present invention, and here, the operation is described specifically for the operation of the control unit 17.
Hereinafter, the operation of the control unit 17 of the in-vehicle speech recognition apparatus according to Embodiment 1 of the present invention will be described in detail with reference to the flowchart of FIG.

The control unit 17 constantly monitors the signal input to the echo cancellation unit 14 and the signal output from the echo cancellation unit 14. That is, the control unit 17 acquires the input / output signal (step ST21), and calculates the ratio of the echo component (echo cancellation amount) in the input signal from the acquired input / output signal (step ST22).
Then, the control unit 17 compares the ratio α of the echo cancellation amount in the input signal with a predefined threshold β, and when the ratio α of the echo cancellation amount exceeds the threshold β (step ST23 “α ≧ β “), The echo cancellation processing by the echo cancellation processing unit 14 is performed (step S24), and if not exceeded (step ST23“ α <β ”), the echo cancellation processing by the echo cancellation processing unit 14 is controlled not to be performed. (Step ST25).

  When the control unit 17 does not cause the echo processing unit 14 to perform echo cancellation processing (step ST25), the control unit 17 waits for the end of the voice recognition processing by the voice recognition unit 15 (step ST26 “Yes”), and echoes The operation mode of the cancel processing unit 14 is switched to the non-operation mode (step ST27), whereby the echo cancellation processing by the echo processing unit 14 can be stopped.

That is, the control unit 17 determines whether or not the speech recognition rate is improved or decreased by removing the echo by the echo canceling unit 14, and based on the determination result, the echo canceling unit 14 performs the echo canceling process. Have switched.
For this reason, according to the vehicle-mounted speech recognition apparatus according to the first embodiment described above, the echo cancellation processing by the echo cancellation unit 14 is stopped when the improvement of the speech recognition rate cannot be expected by the echo removal by the echo cancellation unit 14. As a result, the speech recognition rate can be improved without unnecessary processing load and power consumption. That is, the echo cancellation process can be effectively performed by monitoring the case where the speech recognition rate is improved by the echo cancellation process and the case where the voice recognition rate is decreased due to side effects.

In addition, if the above-mentioned threshold β is determined in advance by experiment or the like to determine whether echo cancellation processing should be performed when the amount of echo cancellation is to be determined by learning from the investigation result, The accuracy of judgment is improved. In particular, when the vehicle type is fixed, it is desirable to determine the threshold value by an experiment with the vehicle type.
Further, according to the first embodiment described above, whether the control unit 17 switches the operation mode of the echo cancellation processing unit 14 is obtained by the echo cancellation processing unit 14 acquiring a determination result from the control unit 17 and switching the operation mode. The same effect can be obtained.

Embodiment 2. FIG.
FIG. 3 is a block diagram showing the internal configuration of the in-vehicle speech recognition apparatus according to Embodiment 2 of the present invention.
The difference from the first embodiment shown in FIG. 1 is that the control unit 17 replaces the echo cancellation amount determination unit 170 to detect whether there is an output signal from the sound source 12 and performs echo cancellation processing according to the result. The output signal presence / absence determination unit 171 that determines whether or not to perform the above-described configuration is provided. Other configurations are the same as those of the first embodiment shown in FIG.

  As described above, according to the in-vehicle speech recognition apparatus according to the second embodiment of the present invention, when there is an output signal from the sound source 11, an improvement in the speech recognition rate can be expected by echo cancellation processing, and when there is no output signal, processing is performed. Since no load or power consumption occurs, waste can be saved.

Embodiment 3 FIG.
FIG. 4 is a block diagram showing the internal configuration of the in-vehicle speech recognition apparatus according to Embodiment 3 of the present invention.
The difference from the first embodiment shown in FIG. 1 is that a signal amplifying unit 18 for amplifying the output signal of the sound source 11 is newly added, and the control unit 17 is replaced with an echo cancellation amount determining unit 170, and a sound volume determining unit 172. And the output level determination unit 173. That is, when the volume setting that is the amplification amount of the signal amplification unit 18 can be detected, or when the output signal level to the sound output unit 12 can be detected, the control unit 17 determines the volume determination unit 172 or the output. The echo cancellation unit 14 is controlled according to the detection result of the level determination unit 173 to determine whether or not to perform the echo cancellation process. Other configurations are the same as those of the first embodiment shown in FIG.

That is, according to the third embodiment, in the control unit 17, the volume determination unit 172 acquires the output signal of the sound source 11 amplified by the signal amplifier 18 to detect the volume setting amount, and the detection result and the threshold value To determine whether or not to perform echo cancellation processing by the echo cancellation processing unit 14. Alternatively, the output level determination unit 173 detects the output signal level to the sound output unit 12, and determines whether or not the echo cancellation processing by the echo cancellation processing unit 14 is performed by comparing the detection result with a threshold value. is doing.
For this reason, according to the on-vehicle speech recognition apparatus according to the third embodiment, when the volume setting or the output signal level to the sound output unit 12 is large, an improvement in the recognition rate can be expected by the echo cancellation processing, and when it is small, the speech recognition is performed. Since an improvement in the rate cannot be expected, by stopping the echo cancellation processing by the echo cancellation unit 14, the speech recognition rate can be improved without any unnecessary processing load and power consumption.

Embodiment 4 FIG.
FIG. 5 is a block diagram showing the internal configuration of the in-vehicle speech recognition apparatus according to Embodiment 4 of the present invention.
The difference from the first embodiment shown in FIG. 1 is that a noise input unit 19 is newly added, and the control unit 17 is replaced with an echo cancellation amount determination unit 170, and an output sound is generated from a signal input to the echo cancellation unit 14. The noise determination unit 174 is configured to detect the magnitude of noise other than the echo due to, and determine whether to perform echo cancellation processing according to the result. Other configurations are the same as those of the first embodiment shown in FIG.

That is, the control unit 17 having the noise determination unit 174 detects the level of noise other than the echo from the signal input to the echo cancellation unit 14, compares the noise level with a predefined threshold, and the result In response to this, it is determined whether or not echo cancellation processing is to be performed by the echo cancellation unit 14, and the echo cancellation unit 14 is controlled.
According to the on-vehicle speech recognition apparatus according to the fourth embodiment, when the speech recognition rate can be improved by the echo cancellation processing according to the noise level, the echo cancellation processing is performed, and the speech recognition rate is improved by the echo cancellation processing. Can be stopped, the processing load and power consumption generated at this time can be reduced.

Embodiment 5 FIG.
FIG. 6 is a block diagram showing an internal configuration of the in-vehicle speech recognition apparatus according to Embodiment 5 of the present invention.
The difference from the first embodiment shown in FIG. 1 is that a vehicle speed sensor 20 is newly added, and the control unit 17 is replaced with an echo cancellation amount determination unit 170, and an echo cancellation process is performed according to a signal output from the vehicle speed sensor 20. The vehicle speed determination unit 175 for determining whether or not to perform the above is provided. Other configurations are the same as those of the first embodiment shown in FIG.

  According to the on-vehicle speech recognition apparatus according to the fifth embodiment, the control unit 17 having the vehicle speed determination unit 175 measures the level of noise with an easy-to-acquire index called vehicle speed. If the vehicle speed is high, the noise level is high. Echo cancellation processing is performed because it can be expected to improve the speech recognition rate, and if it is low, the speech recognition rate cannot be improved. By stopping the echo cancellation processing, it is possible to reduce unnecessary processing load and power consumption. is there.

Embodiment 6 FIG.
FIG. 7 is a block diagram showing the internal configuration of the in-vehicle speech recognition apparatus according to Embodiment 6 of the present invention.

  As shown in FIG. 7, the on-vehicle speech recognition apparatus according to Embodiment 6 of the present invention outputs a sound source 11 that outputs an output sound signal such as music, radio, or guidance sound, and outputs the signal as an output sound. An echo having a sound output unit 12, an audio input unit 13 that outputs the collected sound as an input signal, and an echo cancellation processing function that eliminates an echo of speaker output sound included in the input signal output from the audio input unit 13 A cancel unit 14, a speech recognition unit 15 that performs speech recognition processing on the signal output from the echo cancel unit 14, a result of the speech recognition output from the speech recognition unit 15, and a result thereof are displayed. A recognition result processing unit 16 that executes the function, and a control unit 17 that determines whether or not the speech recognition rate is improved by removing the echo and controls the echo cancellation unit 14; A signal amplifier 18 for outputting amplifies the output signal of the sound source 11, a noise input unit 19, a vehicle speed sensor 20, a navigation device body 21 having a map information or the like has.

The echo cancellation unit 14 can switch whether or not to perform echo cancellation processing according to the determination result of the control unit 17. However, this switching is not performed during the voice recognition processing by the voice recognition unit 15.
The control unit 17 includes an echo cancellation amount determination unit 170, a volume determination unit 172, and an output level determination unit 173. The control unit 17 calculates the ratio of echoes included in the signal output from the voice input unit 13, and calculates the ratio. Is larger than the threshold value, it is determined that there is an effect of echo cancellation by the echo canceling process, and voice recognition using the output of the echo canceling unit 14 is instructed. It is determined that there is a decrease in the speech recognition rate, and an instruction for speech recognition in which echo cancellation processing by the echo cancellation unit 14 is prohibited is issued.

  Note that the echo cancellation amount determination unit 170 is the echo cancellation amount determination unit 170 described in the first embodiment, the volume determination unit 172, and the output level determination unit 173 are the volume determination unit 172 and output described in the second embodiment. The level determination unit 173 is the same as the level determination unit 173. In both cases, the signal input to the echo cancellation unit 14 and the signal output from the echo cancellation unit 14 are monitored, and whether or not the speech recognition rate is improved by removing the echo component. Is used to calculate a parameter (ratio of echo components). Details will be described below.

As a method for the control unit 17 to calculate the ratio of the echo component input to the echo cancellation unit 14, there is a method using an echo cancellation amount, a volume (output level), a noise level, and the like as described above.
When the volume is used, the control unit 17 detects the set volume from the output of the signal amplifying unit 18, and obtains the ratio of the echo component from the volume by calculation. When the output level is used, the control unit 17 detects the level of the output signal input to the sound output unit 12, and obtains the echo ratio from the level by calculation. When the noise level is used, the control unit 17 detects a noise input unit 19 that collects noise separately from the microphone for voice input, a vehicle speed sensor 20, or various types of sensors that detect wiper operation, raindrops, engine speed, and the like. The level of noise and the state of the vehicle are detected from the sensor / navigation device main body 21, and the ratio of noise other than the echo component is obtained by calculation according to the result.

Specifically, when the noise input unit 19 is provided, the control unit 17 detects the direct noise level from the noise input unit 19 and calculates the ratio of noise other than echoes. When the vehicle speed sensor 20 is included, the control unit 17 calculates the ratio of noise other than echoes from the detected vehicle speed. In the case of having a wiper operation or a sensor that detects raindrops, the control unit 17 detects the wiper operation speed or the amount of raindrops, and calculates and determines the influence on noise other than echoes caused by the wiper sound or rain sound. In addition, when the engine speed sensor is provided, the control unit 17 detects the engine speed and obtains the influence on the noise other than the echo caused by the engine sound by calculation.
In addition, when it is possible to acquire the currently traveling road and road conditions from the navigation device main body 21 or the like, the control unit 17 determines whether the vehicle is currently passing through the tunnel or a bad road. And the influence on noise other than echoes caused by tunnels and rough roads can be obtained by calculation.

Subsequently, the control unit 17 integrates the parameters such as the echo cancellation amount, the sound volume (output level), and the noise level detected as described above, and performs weighting calculation according to the priority information defined in advance for each parameter. Then, it is determined whether or not to perform echo cancellation processing.
Here, the priority information is a numerical value assigned to each parameter such as an echo cancellation amount, a volume (output level), a noise level, and the like, and a larger numerical value is in order of the echo cancellation amount, the volume (output level), and the noise level. It is assumed that it is allocated and recorded in a memory built in the control unit 17. The control unit 17 reads out this numerical value for each detected parameter, performs a weighting operation based on this numerical value, finds the sum of at least two detected parameters, compares it with a threshold value, and performs echo cancellation processing according to the magnitude thereof. Whether or not to decide.

FIG. 8 is a flowchart showing the operation of the in-vehicle speech recognition apparatus according to the sixth embodiment of the present invention. Here, the operation is described specifically for the operation of the control unit 17.
Hereinafter, the operation of the control unit 17 of the in-vehicle speech recognition apparatus according to the sixth embodiment of the present invention will be described in detail with reference to the flowchart of FIG.

First, the control unit 17 acquires information necessary for determination for each parameter described above (step ST81), and integrates the determination results (step ST82).
Here, the integration of each determination result means that at least two parameters of the detected echo cancellation level, presence / absence of output signal from sound source, volume setting amount, noise level other than echo, and vehicle speed are at least two parameters. Used, this refers to a weighting operation performed in accordance with priority information defined in advance for each parameter.

Subsequently, the control unit 17 compares γ, which is the result of the weighting calculation, with the threshold δ (step ST83), and if the weighting calculation result γ exceeds the threshold δ (step ST83 “γ ≧ δYes”), The echo canceling process by the cancel unit 14 is performed (step S84), and if it does not exceed (step ST83 “γ <δ”), the echo canceling process by the echo canceling unit 14 is controlled not to be performed (step ST85). .
When the control unit 17 does not cause the echo processing unit 14 to perform echo cancellation processing (step ST85), the control unit 17 waits for the end of the voice recognition processing by the voice recognition unit 15 (step ST86 “Yes”), and echoes The operation mode of the cancel processing unit 14 is switched to the non-operation mode (step ST87), whereby the echo cancellation processing by the echo processing unit 14 can be stopped.

That is, the control unit 17 performs weighting calculation according to priority information defined in advance for each parameter, compares the calculation result with a threshold value, and determines whether or not to perform echo cancellation processing according to the comparison result. is doing.
For this reason, according to the vehicle-mounted speech recognition apparatus according to the sixth embodiment described above, when the improvement of the speech recognition rate cannot be expected due to the echo removal by the echo cancellation unit 14, the echo cancellation processing by the echo cancellation unit 14 is stopped. As a result, the speech recognition rate can be improved without unnecessary processing load and power consumption. Also, with multiple parameters,
Since the presence / absence of the echo cancellation process is determined, the determination accuracy is improved as compared with the first to fifth embodiments.

According to the above-described sixth embodiment of the present invention, whether the control unit 17 switches the operation mode of the echo cancellation processing unit 14, the echo cancellation processing unit 14 acquires the determination result from the control unit 17. The same effect can be obtained even when switching.
Further, in the above-described sixth embodiment, it is configured to determine whether or not to execute the echo cancellation processing based on the determination result by the control unit 17, but whether or not to perform the echo cancellation processing is expressed by a continuous amount, The cancel unit 14 may determine whether or not to perform the echo cancellation processing using a pattern identification method such as a neural network or SVM (Support Vector Machine) using each determination result as a parameter. Note that the threshold value in the above-described pattern identification method can be determined more accurately by investigating in advance by experiment whether or not each parameter should be subjected to echo cancellation processing and by learning from the investigation result. Will improve. In particular, when the vehicle type is fixed, it is preferable to determine the threshold value by an experiment with the vehicle type.

The functions of the echo cancellation processing unit 14, the voice recognition unit 15, and the control unit 17 shown in FIGS. 1 to 7 are all realized by software, or at least a part thereof is realized by hardware. Also good.
For example, the data processing in which the control unit 17 monitors the signal input to the echo cancellation unit, determines whether or not the speech recognition rate is improved by removing the echo component, and controls the echo cancellation processing by the echo cancellation unit May be realized on a computer by one or a plurality of programs, or at least a part thereof may be realized by hardware.

It is a block diagram which shows the internal structure of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows the internal structure of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 2 of this invention. It is a block diagram which shows the internal structure of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 3 of this invention. It is a block diagram which shows the internal structure of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 4 of this invention. It is a block diagram which shows the internal structure of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 5 of this invention. It is a block diagram which shows the internal structure of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 6 of this invention. It is a flowchart which shows operation | movement of the vehicle-mounted speech recognition apparatus which concerns on Embodiment 6 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Vehicle-mounted speech recognition apparatus, 11 Sound source, 12 Sound output part, 13 Voice input part, 14 Echo cancellation part, 15 Voice recognition part, 16 Recognition result processing part, 17 Control part, 18 Signal amplifier, 19 Noise input part, 20 Vehicle speed sensor, 21 navigation device main body, 170 echo cancellation amount determination unit, 171 output signal presence / absence determination unit, 172 volume determination unit, 173 output level determination unit, 174 noise determination unit, 175 vehicle speed determination unit.

Claims (8)

A vehicle-mounted speech recognition device including an echo cancellation unit that performs echo cancellation processing to remove an echo component included in input speech,
With multiple sound sources,
An echo canceling unit that removes an echo component included in the signal, using the input sound collected via the microphone and the signal from the sound source as input, and
A speech recognition unit that recognizes the output of the echo cancellation unit;
A controller that monitors a signal input to the echo cancellation unit, determines whether or not a speech recognition rate is improved by removing an echo component, and controls echo cancellation processing by the echo cancellation unit;
A vehicle-mounted speech recognition apparatus comprising: The controller is
The in-vehicle speech recognition apparatus according to claim 1, wherein an echo cancellation level is detected, and it is determined whether or not to perform echo cancellation processing according to the detected cancellation level. The controller is
3. The method according to claim 1, wherein whether or not there is an output from the sound source is detected, and whether or not echo cancellation processing by the echo cancellation processing unit is performed is determined according to the detection result. In-vehicle voice recognition device. The controller is
Obtaining a signal obtained by amplifying the output of the sound source, detecting a volume setting amount of the signal, and determining whether to perform echo cancellation processing by the echo cancellation processing unit according to the detection result; The in-vehicle voice recognition device according to claim 1 or 2. The controller is
2. The level of noise other than echoes is detected from a signal input to the echo cancellation unit, and it is determined whether or not to perform echo cancellation processing by the echo cancellation unit according to the detection result. The vehicle-mounted speech recognition apparatus according to any one of claims 1 to 4. The controller is
The in-vehicle voice recognition device according to claim 5, wherein a vehicle speed is detected, and it is determined whether or not to perform echo cancellation processing by the echo cancellation unit according to the detection result. The controller is
The echo cancellation processing by the echo processing unit is stopped only when the speech recognition unit is not recognizing the input speech when it is determined that the echo cancellation processing is unnecessary. The in-vehicle voice recognition device according to claim 1. The controller is
At least two parameters among the detected echo cancellation level, presence / absence of output signal from sound source, volume setting amount, noise level other than echo, and vehicle speed are defined in advance for each parameter. The weighting calculation is performed according to the priority information, the calculation result is compared with a predetermined threshold value, and it is determined whether or not an echo cancellation process is performed according to the comparison result. The vehicle-mounted speech recognition device according to claim 1.

Images