JP4425718B2 - Voice recognition device for vehicles - Google Patents

Description

  The present invention relates to a vehicle voice recognition apparatus that recognizes input voice and converts it into information that can be understood by a machine.

Conventionally, a speech recognition device that recognizes input speech and converts it into information that can be understood by the machine is, for example, by adding noise with obvious characteristics similar to the average spectrum of speech to the input speech and adding the noise. Some voices are recognized by collating feature parameters obtained from voice with a standard pattern of voice. Specifically, by actively adding well-known noise (noise close to human voice) to the input voice, environmental noise mixed from the microphone and electrical noise superimposed on the input system (white) Reduce the effects of noise that is not well understood, such as (close to noise). Thereby, even when various environmental noises are mixed into the input voice from the microphone or when the electrical noise of the input circuit is superimposed, it is possible to realize a voice recognition device that can obtain a stable and high recognition rate. (For example, refer to Patent Document 1).
JP-A-6-43892

By the way, in the conventional speech recognition apparatus, there is a problem that consideration is not given to the case where the speech recognition apparatus is mounted on a vehicle such as an automobile. Specifically, when the speech recognition device is mounted on a vehicle such as an automobile, the amount of noise in the passenger compartment differs between when the vehicle is stopped and when the vehicle is running. If there is always a steady noise in the passenger compartment, the added noise is less effective and wasteful even if the noise is always positively added to the input voice as in the conventional voice recognition device. There is a problem that there are many.
On the other hand, the passenger in the passenger seat and the passenger in the passenger seat may make a voice at the same time. For example, when trying to recognize the voice of the passenger in the driver seat, the voice of the passenger in the passenger seat is recognized by the voice of the driver in the driver seat. Since the interior of the vehicle is quiet especially when the vehicle is stopped, it is possible to actively cancel out such unsteady noise as in the case of a conventional voice recognition device. There is also a demand to add stationary noise.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a vehicle voice recognition device capable of recognizing voice efficiently and accurately according to the surrounding environment.

In order to solve the above problems, a vehicle voice recognition apparatus according to the invention of claim 1 includes a voice input unit (for example, a voice input unit 4 in an embodiment described later) capable of inputting by voice and the voice input unit. A voice recognition means (for example, a voice recognition unit 6 or a voice recognition unit 7 in an embodiment to be described later) that performs voice recognition of the input voice and a steady state that determines whether or not the steady noise amount in the vehicle interior is equal to or greater than a predetermined value. When the steady noise amount determining means determines that the steady noise amount in the vehicle interior is smaller than a predetermined value (for example, in-vehicle steady noise determining unit 1 in an embodiment to be described later) and the steady noise amount determining means. A stationary noise addition judging means for judging whether or not stationary noise needs to be added to the voice inputted by (for example, a noise addition necessity judging section 2 in an embodiment to be described later), and input by the voice input means. When the stationary noise addition determining means determines that it is necessary to add stationary noise to the received voice, stationary noise adding means (for example, an embodiment described later) adds the stationary noise to the voice input by the voice input means. A stationary noise generating unit 3 and a stationary noise adding unit 5), and the stationary noise addition determining unit includes an occupant detecting unit (for example, a seating sensor of an embodiment described later) for detecting an occupant in the vehicle interior, Based on the number of occupants in the passenger compartment detected by the occupant detection means or the position of the occupants in the passenger compartment, whether or not it is necessary to add stationary noise to the voice input by the voice input means. It is characterized by judging.

  The vehicle speech recognition apparatus having the above configuration is configured such that when the speech recognition unit executes speech recognition of the speech input by the speech input unit, the steady noise amount is smaller than a predetermined value. If the determination means determines, the steady noise addition determination means determines whether or not the steady noise needs to be added to the voice input by the voice input means. When the stationary noise addition determining unit determines that it is necessary to add stationary noise to the voice input by the voice input unit, the stationary noise adding unit adds the stationary noise to the voice input by the voice input unit. If the steady noise amount in the passenger compartment is smaller than the predetermined value and it is necessary to add the steady noise, the voice recognition of the voice with the steady noise added is executed. In other cases, it is possible to execute voice recognition of the voice as it is inputted by the voice input means.

  According to a second aspect of the present invention, there is provided a vehicular voice recognition apparatus according to the first aspect, wherein the stationary noise addition determining means detects an occupant in the vehicle compartment (for example, described later). A seating sensor of the embodiment), and the noise input to the voice input means based on the number of passengers detected by the passenger detection means or the position of the passengers in the passenger compartment. It is characterized by determining whether it is necessary to add.

Further, stationary noise addition determining means, whether the number of passengers in the cabin, which is detected by the passenger detecting means or on the basis of the occupant position in the vehicle compartment, it is necessary to add the stationary noise in the input speech Therefore, it is possible to predict the occurrence of non-stationary noise that interferes with the voice recognition of the voice of the passenger in the vehicle other than the speaker who should recognize the voice, that is, the voice of the speaker who should recognize the voice, and its influence. The stationary noise can be added to the input voice when necessary.

According to a second aspect of the present invention, there is provided the vehicular voice recognition device according to the first aspect, wherein the stationary noise addition determining means is based on an operating state of an acoustic device provided in the vehicle interior. It is determined whether or not it is necessary to add stationary noise to the voice input by the voice input means.

  In the vehicle speech recognition apparatus having the above configuration, whether or not the steady noise addition determination unit needs to add steady noise to the input voice based on the operating state of the acoustic device provided in the vehicle interior. By predicting the occurrence of non-stationary noise that interferes with the voice recognition of the voice of the speaker that should be recognized, that is, the voice output by the acoustic equipment in the vehicle interior, the voice input when necessary Stationary noise can be added.

According to a third aspect of the present invention, there is provided a vehicular voice recognition apparatus according to the first or second aspect , wherein the voice recognition means stores a plurality of standard patterns related to voice (for example, A standard pattern storage unit 15) of an embodiment to be described later, a standard pattern selection unit (for example, a standard pattern selection unit 16 of an example to be described later) for selecting a standard pattern used for speech recognition from the standard pattern storage unit, Voice collation means (for example, collation unit 14 in an embodiment to be described later) that recognizes voice by collating the voice input by voice input means with the standard pattern selected by the standard pattern selection means; A pattern selection unit selects the standard pattern based on the determination results of the stationary noise amount determination unit and the stationary noise addition determination unit. Characterized in that it-option.

  In the vehicle speech recognition apparatus having the above-described configuration, the standard pattern selection unit included in the speech recognition unit includes a plurality of standard patterns related to speech based on the determination results of the steady noise amount determination unit and the steady noise addition determination unit. The standard pattern used for voice recognition is selected from the standard pattern storage means for storing the voice, and the voice collation means collates the voice input by the voice input means with the standard pattern selected by the standard pattern selection means to obtain the voice. Is recognized based on the determination results of the steady noise amount determination means and the steady noise addition determination means, and whether or not stationary noise is added to the speech is determined, and the standard pattern used for speech recognition is changed to an appropriate one. be able to.

According to the vehicle voice recognition device of claim 1, when the steady noise amount in the vehicle compartment is smaller than the predetermined value and the steady noise needs to be added, the voice of the voice to which the steady noise is added. In other cases, it is possible to execute voice recognition of the voice as it is input by the voice input means.
Therefore, voice recognition is performed by adding stationary noise only when the vehicle interior is quiet and there is little stationary noise, and non-stationary noise that interferes with the speech recognition of the speaker to be recognized is conspicuous. If there is no need to add stationary noise, such as when there is no unsteady noise, or when there is sufficient stationary noise in the passenger compartment, voice recognition is performed without adding unnecessary noise. Thus, there is an effect that it is possible to realize a vehicle voice recognition device that can recognize voice efficiently and accurately according to the environment around the voice recognition device.

Furthermore , based on the number of passengers in the passenger compartment or the position of the passengers in the passenger compartment, the occurrence of non-stationary noise that interferes with the voice recognition of the speaker who should recognize the speech and its effects are predicted and input when necessary. Stationary noise can be added to the generated voice.
Therefore, it is possible to appropriately determine the state of the passenger compartment, add stationary noise to the input voice when necessary, and improve the voice recognition rate.

According to the vehicular voice recognition device according to claim 2 , the occurrence of unsteady noise that interferes with the voice recognition of the voice of the speaker to be voice-recognized is predicted based on the operating state of the acoustic device provided in the vehicle interior. Thus, stationary noise can be added to the input voice when necessary.
Therefore, it is possible to appropriately determine the state of the passenger compartment, add stationary noise to the input voice when necessary, and improve the voice recognition rate.

According to the vehicle voice recognition apparatus of the third aspect , the presence / absence of addition of stationary noise to the voice is determined based on the determination results of the stationary noise amount determination means and the stationary noise addition determination means, and is used for voice recognition. The standard pattern can be changed to an appropriate one.
Therefore, for example, when the vehicle interior is quiet and there is no non-stationary noise that interferes with the speech recognition of the speaker to be recognized, the standard pattern used when the speech recognition is quiet is used. Appropriate standard according to the input voice and the environment around the voice recognition device, such as using a standard pattern created by adding steady noise to voice recognition. The effect of further improving the voice recognition rate using the pattern and realizing the vehicle voice recognition device capable of recognizing the voice efficiently and accurately is obtained.

  Embodiments of the present invention will be described below with reference to the drawings.

(overall structure)
FIG. 1 is a block diagram showing the configuration of a vehicle speech recognition apparatus according to a first embodiment of the present invention.
In FIG. 1, the vehicle speech recognition apparatus according to the present embodiment includes an in-vehicle steady noise determination unit 1 that determines the amount of steady noise generated in a vehicle compartment in a vehicle equipped with the vehicle speech recognition apparatus, and an in-vehicle steady noise determination. Depending on the judgment result of part 1 and the presence or absence of passengers and the presence or absence of non-stationary noise that hinders speech recognition of other people other than the speaker who should be recognized by speech estimation based on the number of passengers, etc. And a noise addition necessity determination unit 2 that determines whether or not it is necessary to add clear stationary noise.

  In addition, the vehicle speech recognition apparatus according to the present embodiment includes a stationary noise generating unit 3 that generates stationary noise with a clear property to be added to input speech, a microphone that acquires a speech, and the like. Based on the determination result of the input unit 4, the noise addition necessity determination unit 2, the stationary noise addition unit 5 that adds the stationary noise generated by the stationary noise generation unit 3 to the voice input by the voice input unit 4, A speech recognition unit 6 that recognizes speech by comparing the speech output from the noise addition unit 5 with a standard pattern stored in advance is provided.

  The speech recognition unit 6 will be described in more detail. The speech recognition unit 6 analyzes the speech output from the stationary noise addition unit 5 using, for example, linear prediction analysis, and the output of the analysis unit 11. For example, a feature parameter extraction unit 12 that obtains, for example, an LPC (linear prediction) cepstrum coefficient as a feature parameter of speech, and a speech standard pattern for comparison with the feature parameter output from the feature parameter extraction unit 12 are stored in advance. The standard pattern storage unit 13 is provided.

  In addition, the speech recognition unit 6 performs pattern matching between the time series data of the feature parameters output from the feature parameter extraction unit 12 and the standard pattern of speech output from the standard pattern storage unit 13, and the degree of similarity with the feature parameters is maximized. And a collation unit 14 that outputs a voice corresponding to the standard pattern as a voice recognition result. Note that the standard pattern of speech stored in the standard pattern storage unit 13 is created in advance using standard pattern creation data for each speech to be recognized.

(Voice recognition processing)
Next, with reference to the drawings, the voice recognition processing of the vehicle voice recognition apparatus of the present embodiment will be described. FIG. 2 is a flowchart showing the voice recognition processing operation of the vehicle voice recognition apparatus of the present embodiment.
In FIG. 2, when the user of the vehicle speech recognition apparatus of this embodiment presses the talk switch and inputs a voice in order to speak (step S <b> 1), the vehicle interior steady noise determination unit 1 The steady noise amount is measured, and it is determined whether or not the steady noise amount in the passenger compartment is equal to or greater than a predetermined value (step S2). Specifically, for example, the steady noise amount in the passenger compartment when the talk switch is pressed and the user's utterance is started is measured, and the S / N ratio between the user's voice and the stationary noise in the passenger compartment is determined. It is determined whether or not the threshold value TH1 is equal to or less than the threshold value TH1, and when the S / N ratio between the user's voice and the steady noise in the vehicle interior is equal to or less than the threshold value TH1, the steady noise amount in the vehicle interior is predetermined. Determined to be greater than or equal to the value. Note that it may be directly determined whether or not the traveling speed Vs of the vehicle is equal to or greater than a predetermined value TH2.

  In step S2, if the vehicle interior steady noise determination unit 1 determines that the amount of steady noise in the vehicle interior is not equal to or greater than a predetermined value (NO in step S2), then the noise addition necessity determination unit 2 performs the voice input unit. 4 determines whether it is necessary to add stationary noise with a clear nature to the voice input in step 4 (step S3). Specifically, the noise addition necessity determination unit 2 detects the number of passengers in the passenger compartment using a passenger detection means such as a seating sensor installed in a vehicle seat, for example, and a speech to be recognized. Whether it is necessary to add stationary noise with obvious characteristics to the input speech based on the number of passengers (passengers) other than passengers (users of the vehicle voice recognition device) and the positions of passengers in the passenger compartment Determine whether.

  For example, if there is an occupant other than the user of the vehicle voice recognition device, the voice of another person other than the speaker who should recognize the voice is input as non-stationary noise that interferes with the voice recognition of the voice of the speaker who should recognize the voice. In order to cancel this, it is determined that it is necessary to add stationary noise with obvious properties.

  Even in such a case, for example, a microphone (speech input unit 4) of a vehicle speech recognition device is installed on an instrument panel or a center console in front of the passenger compartment, and a speaker (a vehicle speech recognition device When a passenger other than the user is present in the rear seat far from the microphone (speech input unit 4) of the vehicle speech recognition device, it can be estimated that the influence of non-stationary noise that disturbs speech recognition is small. The necessity determination unit 2 can determine that it is not necessary to add stationary noise to the input voice. When it is determined whether it is necessary to add stationary noise to the input voice based on the position of the passenger in the vehicle compartment, the microphone (voice input unit 4) of the actual vehicle voice recognition device is installed. Judging from the position and the position of the occupant.

  Further, the noise addition necessity determination unit 2 determines whether or not it is necessary to add stationary noise to the input voice based on the number of voice speakers simultaneously input to the voice input unit 4. Also good. That is, even if there is an occupant other than the speaker (user of the vehicle speech recognition device) that should be speech-recognized, if the occupant is not speaking, the speech recognition of the speaker to be speech-recognized. There is no non-stationary noise that interferes. Therefore, when it is possible to recognize that the number of voice speakers simultaneously input to the voice input unit 4 is one, the noise addition necessity determination unit 2 determines that it is not necessary to add stationary noise to the input voice. When the number of voice speakers simultaneously input to the voice input unit 4 can be recognized as plural, the noise addition necessity determination unit 2 determines that it is necessary to add stationary noise to the input voice.

  In the above description, the source of non-stationary noise that interferes with speech recognition has been described as an occupant other than a speaker (a user of a vehicle speech recognition device) that should be speech-recognized. Since the output sound of the acoustic equipment provided in the vehicle interior can also be considered as one of the non-stationary noises that disturb the speech recognition, the noise addition necessity determination unit 2 adds the steady noise to the input sounds. Whether or not it is necessary to do so may be determined based on the operating state of the audio equipment provided in the passenger compartment. Therefore, when the acoustic device provided in the passenger compartment is not operating, the noise addition necessity determination unit 2 determines that it is not necessary to add stationary noise to the input voice. Moreover, when the audio equipment provided in the vehicle interior is operating, the noise addition necessity determination unit 2 determines that it is necessary to add stationary noise to the input voice.

  The noise addition necessity determination unit 2 determines whether or not it is necessary to add stationary noise to the input voice. The number or position of passengers other than the speaker to be voice-recognized or the voice input unit 4 Judgment may be made based only on the condition of the number of voice speakers that are input simultaneously or the operating state of the acoustic equipment provided in the vehicle interior, or a combination of these conditions.

  On the other hand, when the noise addition necessity determination unit 2 determines in step S3 that it is necessary to add stationary noise to the voice input by the voice input unit 4 (YES in step S3), the stationary noise addition unit 5 Stationary noise with obvious properties generated by the stationary noise generator 3 is added to the voice input by the voice input unit 4 (step S4). Specifically, as shown in FIG. 3 showing the stationary noise to be added to the input voice, the stationary noise as shown in FIG. 3B is added to the input voice signal shown in FIG. Thus, the influence of the non-stationary noise in the A part or B part of the signal shown in FIG.

If the stationary noise can be added to the speech input by the speech input unit 4 by the stationary noise addition unit 5, the analysis unit 11 performs linear prediction analysis, and the feature parameter extraction unit 12 inputs the speech. Feature parameters are extracted (step S5).
When the speech feature parameters can be extracted, the matching unit 14 performs pattern matching between the feature parameter time-series data output from the feature parameter extraction unit 12 and the speech standard pattern output from the standard pattern storage unit 13. (Step S6), the voice corresponding to the standard pattern that maximizes the similarity to the feature parameter is output as the voice recognition result (step S7), and the voice recognition process is terminated.

  In step S2, if the in-vehicle steady noise determining unit 1 determines that the amount of steady noise in the vehicle interior is equal to or greater than a predetermined value (YES in step S2), the steady noise is not added to the input voice. In step S5, the analysis unit 11 performs linear prediction analysis of the speech input by the speech input unit 4 as it is, and the feature parameter extraction unit 12 extracts the feature parameters of the input speech (step S5).

  In step S3, when the noise addition necessity determination unit 2 determines that it is not necessary to add stationary noise to the voice input by the voice input unit 4 (NO in step S3), the input voice is steady. In step S5, noise is not added, and the analysis unit 11 performs linear prediction analysis of the speech input by the speech input unit 4 as it is, and the feature parameter extraction unit 12 extracts the feature parameters of the input speech (step S5). S5).

  As described above, in the vehicle speech recognition apparatus according to the present embodiment, when the in-vehicle steady noise determination unit 1 determines that the amount of steady noise in the vehicle interior is not equal to or greater than a predetermined value, the noise addition necessity determination unit 2 The voice input unit 4 is determined based on the number and positions of passengers other than the speakers to be recognized by voice, the number of voice speakers simultaneously input to the voice input unit 4, or the operating state of the acoustic equipment provided in the vehicle interior. To determine whether it is necessary to add stationary noise with a clear nature to the input voice. If it is determined that it is necessary to add stationary noise to the input speech, the stationary noise adding unit 5 adds the stationary noise that is clearly generated by the stationary noise generating unit 3 to the input speech. To do. Next, the analysis unit 11 performs linear prediction analysis, and the feature parameter extraction unit 12 extracts the feature parameters of the input speech. Then, the matching unit 14 performs pattern matching between the time series data of the feature parameter output from the feature parameter extraction unit 12 and the standard pattern of the voice output from the standard pattern storage unit 13, and the similarity with the feature parameter is maximized. The voice corresponding to the standard pattern is output as the voice recognition result.

Thereby, when the steady noise amount in the passenger compartment is smaller than the predetermined value and it is necessary to add the steady noise, the voice recognition of the voice to which the steady noise is added is executed, and in other cases, Voice recognition of the voice as it is input by the voice input unit 4 can be executed.
Therefore, for example, when the vehicle is stationary, the vehicle interior is quiet and there is little steady noise, and only when there is noticeable non-stationary noise that interferes with voice recognition of the speaker's voice, such as the passenger's voice. Recognize voice by adding stationary noise, for example, when the vehicle interior is quiet and no non-stationary noise is generated, or when sufficient stationary noise is generated because the vehicle is running, etc. Voice for vehicles that can recognize voice efficiently and accurately according to the surrounding environment of the voice recognition device, such as recognizing voice without adding noise unnecessarily when it is not necessary to add stationary noise The effect that a recognition device can be realized is obtained.

Further, the noise addition necessity determination unit 2 determines the number and position of passengers other than the speaker to be recognized by voice, the number of voice speakers input simultaneously to the voice input unit 4, or the sound provided in the passenger compartment. By determining whether or not it is necessary to add stationary noise with obvious characteristics to the voice input by the voice input unit 4 based on the operating state of the equipment, the passenger's voice and the acoustic equipment in the passenger compartment Voice that is input when necessary by predicting the occurrence and effects of non-stationary noise that interferes with the voice recognition of the speaker's voice that should be recognized, such as the output voice Stationary noise can be added to.
Therefore, it is possible to improve the speech recognition rate by more accurately determining the state to which stationary noise should be added and adding stationary noise to the input speech.

Next, a second embodiment of the present invention will be described.
(overall structure)
FIG. 4 is a block diagram showing the configuration of the vehicle voice recognition apparatus according to the second embodiment of the present invention.
In FIG. 4, the vehicle speech recognition apparatus according to the present embodiment includes a vehicle interior steady noise determination unit 1, a noise addition necessity determination unit 2, a steady noise generation unit 3, a speech input unit 4, and a steady noise addition unit 5. And a voice recognition unit 7. Here, the in-vehicle steady noise determining unit 1, the noise addition necessity determining unit 2, the steady noise generating unit 3, the voice input unit 4, and the steady noise adding unit 5 are the first of the present invention shown in FIG. Since it is the same as the component which comprises the speech recognition apparatus for vehicles of an Example, description is abbreviate | omitted.

  On the other hand, the speech recognition unit 7 will be described in more detail. The speech recognition unit 7 analyzes the speech output from the stationary noise addition unit 5 using, for example, linear prediction analysis, and the output of the analysis unit 11. For example, a feature parameter extraction unit 12 that obtains, for example, an LPC (linear prediction) cepstrum coefficient as a feature parameter of speech, and a speech standard pattern for comparison with the feature parameter output from the feature parameter extraction unit 12 are stored in advance. The standard pattern storage unit 15 is provided.

  In addition, the standard pattern of the voice stored in advance in the standard pattern storage unit 15 is created in advance using standard pattern creation data that does not change anything for each voice to be recognized. And for each voice to be recognized, created in advance using standard pattern creation data obtained by adding the same quality of stationary noise generated by the stationary noise generating unit 3 at a constant rate, Prepare multiple types of standard patterns.

  The voice recognition unit 7 also includes a plurality of types of standards stored in the standard pattern storage unit 15 in the standard pattern storage unit 15 based on the determination results of the in-vehicle steady noise determination unit 1 and the noise addition necessity determination unit 2. A standard pattern selection unit 16 for instructing which standard pattern in the pattern is used for speech recognition, time series data of feature parameters output from the feature parameter extraction unit 12, and a standard pattern selection unit 16 based on the selection A matching unit 14 is provided that performs pattern matching with a standard pattern of speech output from the pattern storage unit 15 and outputs speech corresponding to a standard pattern that maximizes the similarity to the feature parameter as a speech recognition result.

(Voice recognition processing)
Next, with reference to the drawings, the voice recognition processing of the vehicle voice recognition apparatus of the present embodiment will be described. FIG. 5 is a flowchart showing the voice recognition processing operation of the vehicle voice recognition apparatus of the present embodiment.
In FIG. 5, when the user of the vehicle speech recognition apparatus of the present embodiment presses the talk switch and inputs a voice in order to speak (step S11), the in-vehicle steady noise determination unit 1 The steady noise amount is measured, and it is determined whether or not the steady noise amount in the passenger compartment is a predetermined value or more (step S12). Note that the determination method of the steady noise amount in the vehicle interior in step S12 is the same as the processing in step S2 of the vehicle speech recognition apparatus of the first embodiment.

  In step S12, if the vehicle interior steady noise determination unit 1 determines that the amount of steady noise in the vehicle interior is not greater than or equal to a predetermined value (NO in step S12), then the noise addition necessity determination unit 2 performs the voice input unit. 4 determines whether it is necessary to add stationary noise with a clear nature to the voice input in step 4 (step S13). Note that the method for determining whether or not stationary noise needs to be added to the input speech in step S13 is the same as the processing in step S3 of the vehicle speech recognition apparatus of the first embodiment.

Also, in step S13, when the noise addition necessity determination unit 2 determines that it is necessary to add stationary noise to the voice input by the voice input unit 4 (YES in step S13), FIG. As described with reference to FIG. 3, the stationary noise adding unit 5 adds the stationary noise that is clearly generated by the stationary noise generating unit 3 to the voice input by the voice input unit 4 (step S14).
If the stationary noise can be added to the speech input by the speech input unit 4 by the stationary noise addition unit 5, the analysis unit 11 performs linear prediction analysis, and the feature parameter extraction unit 12 inputs the speech. Feature parameters are extracted (step S15).

  In step S12, when the vehicle interior stationary noise determination unit 1 determines that the amount of stationary noise in the vehicle interior is equal to or greater than a predetermined value (YES in step S12), the stationary noise is not added to the input voice. In step S15, the analysis unit 11 performs the linear prediction analysis of the speech input by the speech input unit 4 as it is, and the feature parameter extraction unit 12 extracts the feature parameter of the input speech (step S15).

  On the other hand, when the stationary noise is added to the voice input by the voice input unit 4 in step S14, or when it is determined in step S12 that the stationary noise amount in the vehicle interior is equal to or greater than the predetermined value, the stationary noise is added. Therefore, the standard pattern selection unit 16 stores the standard pattern storage unit 15 in the standard pattern storage unit 15 based on the determination result of the in-vehicle steady noise determination unit 1 or the noise addition necessity determination unit 2. A standard pattern 1 created from standard pattern creation data obtained by adding, at a constant rate, the same quality as the stationary noise generated by the stationary noise generating unit 3 from a plurality of types of standard patterns stored in the pattern storage unit 15 Is used for voice recognition (step S16).

  Then, after extracting the speech feature parameters and designating the speech standard pattern to be used for speech recognition, the collating unit 14 selects the feature parameter time-series data output from the feature parameter extracting unit 12 and the standard pattern selection. Based on the selection of the unit 16, pattern matching with the standard pattern 1 of the voice output from the standard pattern storage unit 15 is performed (step S17), and the voice corresponding to the standard pattern having the maximum similarity with the feature parameter is recognized as voice. As a result (step S18), the speech recognition process is terminated.

  In Step S13, when the noise addition necessity determination unit 2 determines that it is not necessary to add stationary noise to the voice input by the voice input unit 4 (NO in Step S13), the stationary noise is added to the input voice. The analysis unit 11 performs the linear prediction analysis of the speech input by the speech input unit 4 as it is, and the feature parameter extraction unit 12 extracts the feature parameter of the input speech (step S19).

  If it is determined that it is not necessary to add stationary noise to the voice input by the voice input unit 4 in step S13, the voice to which the stationary noise is not added is subject to voice recognition. Based on the determination result of the noise addition necessity determination unit 2, the unit 16 causes the standard pattern storage unit 15 to change the standard from among a plurality of types of standard patterns stored in the standard pattern storage unit 15. The standard pattern 2 created from the pattern creation data is instructed to be used for voice recognition (step S20).

  Then, after extracting the speech feature parameters and designating the speech standard pattern to be used for speech recognition, the collating unit 14 selects the feature parameter time-series data output from the feature parameter extracting unit 12 and the standard pattern selection. Based on the selection of the unit 16, pattern matching with the standard pattern 2 of the voice output from the standard pattern storage unit 15 is performed (step S21), and the voice corresponding to the standard pattern that maximizes the similarity with the feature parameter in step S18 Is output as a voice recognition result (step S18), and the voice recognition process is terminated.

  As described above, the vehicular speech recognition apparatus of the present embodiment is similar to the vehicular speech recognition apparatus of the first embodiment in that the steady noise amount in the vehicle interior 1 is equal to or greater than a predetermined value. When it is determined that there is no noise, the noise addition necessity determination unit 2 determines whether it is necessary to add stationary noise with a clear nature to the voice input by the voice input unit 4. If it is determined that it is necessary to add stationary noise to the input speech, the stationary noise adding unit 5 adds the stationary noise that is clearly generated by the stationary noise generating unit 3 to the input speech. To do. Next, the analysis unit 11 performs linear prediction analysis, and the feature parameter extraction unit 12 extracts the feature parameters of the input speech. On the other hand, if it is determined that it is not necessary to add stationary noise to the input speech, the stationary noise is not added to the input speech, and the analysis unit 11 performs linear prediction analysis of the input speech as it is. The feature parameter extraction unit 12 extracts the feature parameters of the input speech.

  In the vehicle speech recognition apparatus according to the present embodiment, when it is determined that it is necessary to add stationary noise to the input voice, the standard pattern selection unit 16 makes a stationary noise to the standard pattern storage unit 15. An instruction is given to use the standard pattern 1 created from the data for creating the standard pattern in which the same quality as the stationary noise generated by the generating unit 3 is added at a constant rate for speech recognition. On the other hand, when it is determined that it is not necessary to add stationary noise to the input voice, an instruction is given to use the standard pattern 2 created from the standard pattern creation data without any modification for voice recognition. . Then, the matching unit 14 performs pattern matching between the time series data of the feature parameter output from the feature parameter extraction unit 12 and the standard pattern 1 of the voice output from the standard pattern storage unit 15, and the degree of similarity with the feature parameter is maximized. The voice corresponding to the standard pattern is output as the voice recognition result.

Thereby, when the steady noise amount in the passenger compartment is smaller than the predetermined value and it is necessary to add the steady noise, the voice recognition of the voice to which the steady noise is added is executed, and in other cases, While performing speech recognition of the speech as it is input by the speech input unit 4, it is determined whether or not stationary noise is added to the speech based on the determination results of the in-vehicle steady noise determining unit 1 and the noise addition necessity determining unit 2. Thus, the standard pattern used for speech recognition can be changed to an appropriate one.
Therefore, as in the first embodiment, when the vehicle interior is quiet and unsteady noise that disturbs the voice recognition of the speaker to be recognized, such as the passenger's voice, is noticeable, the stationary noise is added. This is unnecessary when there is no need to add steady noise, such as when there is no steady noise or when there is sufficient steady noise because the vehicle is running. When recognizing speech without adding noise and when steady noise is not added to the speech, use a standard pattern that is used when the speech is quiet, and when stationary noise is added to the speech Uses a standard pattern created by adding stationary noise to speech recognition and uses an appropriate standard pattern according to the input speech and the surrounding environment of the speech recognition device. Improve Effect that efficiently and accurately can be realized recognizable speech recognition device for a vehicle audio.

It is a block diagram which shows the structure of the speech recognition apparatus for vehicles of 1st Example of this invention. It is a flowchart which shows the speech recognition process operation | movement of the speech recognition apparatus for vehicles of the Example. It is a figure which shows the stationary noise added with the input audio | voice. It is a block diagram which shows the structure of the speech recognition apparatus for vehicles of 2nd Example of this invention. It is a flowchart which shows the speech recognition process operation | movement of the speech recognition apparatus for vehicles of the Example.

Explanation of symbols

1 Car interior steady noise judgment section (steady noise amount judgment means)
2 Noise addition necessity judgment section (stationary noise addition judgment means)
3 Stationary noise generator (stationary noise addition means)
4 Voice input part (voice input means)
5 Stationary noise addition unit (stationary noise addition means)
6, 7 Voice recognition unit (voice recognition means)
14 Verification unit (voice verification means)
15 Standard pattern storage (standard pattern storage means)
16 Standard pattern selection unit (standard pattern selection means)

Claims (3)

Voice input means capable of voice input;
Voice recognition means for executing voice recognition of the voice input by the voice input means;
Steady noise amount determination means for determining whether or not the steady noise amount in the passenger compartment is equal to or greater than a predetermined value;
When the steady noise amount determining means determines that the steady noise amount in the passenger compartment is smaller than a predetermined value, it is determined whether or not the steady noise needs to be added to the voice input by the voice input means. Stationary noise addition determination means;
Stationary noise addition means for adding stationary noise to the voice input by the voice input means when the stationary noise addition judgment means determines that it is necessary to add stationary noise to the voice input by the voice input means It equipped with a door,
The stationary noise addition determining means is
Occupant detection means for detecting an occupant in the passenger compartment,
Based on the number of occupants in the passenger compartment detected by the occupant detection means or the position of the occupants in the passenger compartment, whether or not it is necessary to add stationary noise to the voice input by the voice input means. A vehicle speech recognition apparatus characterized by determining . The stationary noise addition determining means determines whether or not it is necessary to add stationary noise to the voice input by the voice input means, based on the operating state of the acoustic device provided in the vehicle interior. The vehicular voice recognition device according to claim 1, wherein The voice recognition means
Standard pattern storage means for storing a plurality of standard patterns related to speech;
Standard pattern selection means for selecting a standard pattern used for speech recognition from the standard pattern storage means;
Voice collating means for recognizing voice by collating the voice input by the voice input means and the standard pattern selected by the standard pattern selecting means;
The standard pattern selecting means, on the basis of the determination result of the stationary noise amount determining means and the stationary noise addition determining unit, a vehicle for speech recognition according to claim 1 or 2, characterized in that selects the reference pattern apparatus.

Images