JP2009251470A - In-vehicle information system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an in-vehicle information system with improved speech recognition precision by learning the feature of speech of each user. <P>SOLUTION: This in-vehicle information system includes: a first input means to input instructions manually, a second input means to input voice, a speech recognition means to recognize the voice input into the second input means, an information processing means to output predetermined information based on the contents of the instructions input into the first input means or the semantic contents of the voice recognized by the speech recognition means, and a learning means to learn a recognition method of the speech recognition means based on data indicating the contents of the instructions and voice data indicating the voice input into the second input means if the instructions are input into the first input means within a predetermined time period after the voice is input in the second input means. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an in-vehicle information system in which the accuracy of speech recognition is improved by learning features of a user's utterance.

  2. Description of the Related Art Conventionally, various ingenuity has been made in an in-vehicle navigation device in order to improve recognition accuracy in voice input.

For example, if a misrecognition occurs during voice recognition and the error is recognized even after multiple redoes, replace the recognition template with another template and re-execute the speech recognition. A navigation device has been proposed (see, for example, Patent Document 1).
JP 2002-108386 A

  By the way, in the conventional navigation apparatus as described above, since the template used for speech recognition is prepared assuming an unspecified number of users, there is a limit in improving accuracy even if it is replaced with another template. .

  SUMMARY OF THE INVENTION An object of the present invention is to provide an in-vehicle information system in which speech recognition accuracy is improved by learning features of individual users' utterances.

  An in-vehicle information system according to one aspect of the present invention includes a first input unit that inputs a command by manual operation, a second input unit that inputs a voice, and a voice recognition that recognizes a voice input to the second input unit. Means, information processing means for outputting predetermined information based on the content of the command input to the first input means, or the meaning content of the voice recognized by the voice recognition means, and the second input means When a command is input to the first input means within a predetermined time after the voice is input, based on data representing the content of the command and voice data representing the voice input to the second input means. Learning means for learning a recognition method in the voice recognition means.

  Further, the learning means learns the characteristics of the voice input to the second input means based on the degree of coincidence between the data representing the content of the command input to the first input means and the voice data. Thus, the recognition method in the voice recognition means may be learned.

  The voice recognition dictionary further stores a reference value for evaluating the feature of the voice and a threshold value for specifying the semantic content of the voice data, and the voice recognition means is input to the second input means. The speech content evaluation value and the reference value stored in the speech recognition dictionary are compared with the threshold value to recognize the semantic content of the speech, and the learning means The recognition method in the voice recognition unit may be learned by changing the threshold based on the degree of coincidence between the data representing the content of the command input to the first input unit and the voice data.

  The voice feature may be at least one of a voice print, an accent, a pitch, and an utterance content.

  The first input means may be a touch panel display means or a remote operation device.

  The information processing means may be included in arithmetic processing means of the navigation device.

  According to the present invention, it is possible to provide a unique effect that an in-vehicle information system with improved voice recognition accuracy can be provided by learning the features of each user's utterance.

  Embodiments to which the in-vehicle information system of the present invention is applied will be described below.

  FIG. 1 is a diagram showing a configuration of an in-vehicle information system according to the present embodiment. This in-vehicle information system is a navigation device 10, and a learning function described below is realized by a navigation ECU (Electronic Control Unit) 11.

  In addition to the navigation ECU 11, the navigation device 10 includes a touch panel 12, a current position detection unit 13, an orientation detection unit 14, a route search unit 15, a map database 16, a microphone 17, a voice recognition dictionary 18, a reception unit 19, and a remote operation. A device 20 is provided.

  The navigation ECU 11 is configured around a microcomputer including a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like via a bus (not shown).

  In addition to the electronic map and the position of the host vehicle, the touch panel 12 displays input switches for realizing predetermined functions (for example, a route search function) of the navigation device 10. This touch panel 12 is an input means for inputting a command by manual operation. For example, the touch panel 12 may be a combination of a liquid crystal monitor and a matrix switch. When the input switch displayed on the panel is pressed by the user, What is necessary is just to be comprised so that operation is possible.

  The current position detection unit 13 is composed of a GPS receiver that calculates the current position and traveling speed of the vehicle using a satellite navigation system (GPS).

  The direction detection unit 14 is configured by a gyrocompass.

  The route search unit 15 is configured to search for an optimum route from candidate routes by weighting search conditions designated by the user.

  Moreover, the map database 16 should just be able to store the electronic map required for a navigation apparatus, for example, is comprised with a hard disk.

  The microphone 17 is an input means for inputting sound. The input sound is converted into sound data and input to the navigation ECU 11. The navigation ECU 11 recognizes the meaning content of the voice data by a voice recognition function described later.

  The voice recognition dictionary 18 is a dictionary used when the navigation ECU 11 executes the voice recognition function, and is used for a reference value as a reference for evaluating voice data of various voices and a determination for specifying the voice data. A database that stores threshold values and semantic content data representing the semantic content of the speech in association with each other and stores them as group data, for example, a hard disk. In addition, the threshold value used for the determination for specifying audio | voice data is mentioned later.

  The receiving unit 19 is a receiving unit that receives a command transmitted from the remote control device 20, and is configured to receive a command input to the remote control device 20 by a manual operation and transmit the command to the navigation ECU 11.

  The navigation device 10 is configured such that a user can input a command by manual operation through the touch panel 12 or the remote device 20, and can input a command by voice through the microphone 17, and the above-described detection unit (13, 14, It is only necessary to be configured so that desired guidance information can be provided to the user by deriving position data representing the position of the host vehicle on the electronic map using the information detected in 15) and map data.

  In the present embodiment, the navigation ECU 11 has a learning function for learning a recognition method in the voice recognition function based on the features of the user's utterance, in addition to the guidance function for providing desired guidance information to the user. . This learning function will be described with reference to FIG.

  FIG. 2 is a block diagram showing the learning function of the speech recognition method in the in-vehicle information system of the present embodiment.

  The navigation ECU 11 includes an input management unit 21, an input analysis control unit 22, a manual input analysis unit 23, a speech recognition unit 24, and a speech recognition learning unit 25. FIG. 2 shows only the blocks necessary for realizing the speech recognition function and the learning function of the recognition method among the functions included in the navigation ECU 11. The navigation ECU 11 provides the user with desired guidance information. It includes other blocks for realizing a guidance function for the purpose.

  The input management unit 21 manages commands input to the touch panel 12, the microphone 17, or the remote operation device 20. The input command is transmitted to the input analysis control unit 22.

  The input analysis control unit 22 has a function of analyzing an input command, and transmits the command input from the input management unit 21 to the manual input analysis unit 23 or the voice recognition unit 24. A command input from the touch panel 12 and the remote device 20 is transmitted to the manual input analysis unit 23, and a command input from the microphone 17 is converted into voice data and transmitted to the voice recognition unit 24.

  In addition, when an instruction is input from the microphone 17 via the input management unit 21, the input analysis control unit 22 counts the elapsed time thereafter. In this case, when a command is input from the touch panel 12 or the remote operation device 20 via the input management unit 21 within a predetermined time, a signal indicating that is transmitted to the speech recognition learning unit 25. The predetermined time is set to 5 seconds, for example.

  The manual input analysis unit 23 analyzes commands input from the touch panel 12 and the remote device 20. A command input from the touch panel 12 and the remote device 20 is a command for operating the navigation device 10. For example, when the “current location” button on the touch panel 12 or the remote device 20 is pressed in order to input the current location, a command for selecting a mode for inputting the current location is used. The manual input analysis unit 23 transmits this command to the speech recognition learning unit 25 as necessary.

  The voice recognition unit 24 recognizes the meaning content of the voice data transmitted from the input analysis control unit 22 using the voice recognition dictionary 18. This recognition process is performed by evaluating a plurality of authenticators included in the voice data. The authenticator is a voiceprint, accent, pitch, and utterance content. Among these, the voiceprint, accent, and pitch represent the characteristics of the utterance.

  This evaluation is performed by evaluating voice data evaluation values (voice print evaluation value, accent evaluation value, pitch evaluation value, and utterance content evaluation value) and evaluation reference parameters (voice print parameter Pa, accent parameter Pb, pitch use). This is done by determining the similarity between the parameter Pc and the speech content parameter Pd). Data representing values (reference values) of these parameters (voice print parameter Pa, accent parameter Pb, pitch parameter Pc, and utterance content parameter Pd) are stored in the speech recognition dictionary 18.

  The evaluation of the voiceprint and the utterance content is performed, for example, by evaluating words and phonemes included in the voice data using a Hidden Markov Model (HMM). For this evaluation, a voiceprint parameter Pa and a speech content parameter Pd are used as evaluation criteria.

  Accent evaluation is realized by performing authentication using a voice pitch distribution represented by voice data. In this evaluation, an accent parameter Pb is used as an evaluation criterion.

  The pitch evaluation is realized by performing authentication using the voice pitch represented by the voice data. In this evaluation, a pitch parameter Pc is used as an evaluation criterion.

  The similarity is determined by evaluating each evaluation value (voice print evaluation value, accent evaluation value, pitch evaluation value, and utterance content evaluation value) and evaluation reference parameters (voice print parameter Pa, accent parameter Pb, pitch parameter). Whether the degree of similarity with each value (reference value) of Pc and utterance content parameter Pd) is equal to or greater than each of voice print threshold Ta, accent threshold Tb, pitch threshold Tc, and utterance content threshold Td Is done by. If all the similarities are greater than or equal to the threshold, it is determined that they are similar, and if any of the similarities is less than the threshold, it is determined that they are dissimilar.

  FIG. 3 is a diagram showing a data structure of the speech recognition dictionary 18 in the in-vehicle information system of the present embodiment. In this way, identification IDs (Identification) are assigned to various words in the speech recognition dictionary, and for each identification ID, voiceprint threshold Ta, accent threshold Tb, pitch threshold Tc, utterance content threshold Td, and semantic content. An ID is stored in association with each other.

  Note that the data shown in FIG. 3 is a part of the data stored in the voice recognition dictionary 18, and actually data of various words is stored. Further, although not shown in FIG. 3, for each identification ID, parameters (voice print parameter Pa, accent parameter Pb, pitch parameter Pc, and utterance content parameter Pd) serving as an average evaluation standard for the word are included. Data representing a value (reference value) is also stored.

  For example, the word with the identification ID “0001” is a word represented by a voiceprint threshold Ta of 0.8, an accent threshold Tb of 0.9, a pitch threshold Tc of 0.7, and an utterance content threshold Td of 0.75. The meaning of the words satisfying these parameters is “M00001” as the semantic content ID.

  Here, if the semantic content data corresponding to the semantic content ID “M00001” represents “current location”, the recognition method is as follows.

  The voice recognition unit 24 uses the voiceprint parameter Pa, the accent parameter Pb, the pitch parameter Pc, and the speech content parameter Pd to evaluate the voice data evaluation value (voiceprint evaluation value, accent evaluation value, pitch evaluation value, and The similarity to the utterance content evaluation value) is calculated.

  The voice recognizing unit 24 determines that the calculated voiceprint similarity, accent similarity, pitch similarity, and speech content similarity are all the voiceprint threshold Ta, accent threshold Tb, pitch threshold Tc, and speech content threshold Td. If it is equal to or greater than the value, it is determined that the voice data and the word with the identification ID “0001” are similar.

  The voice recognition unit 24 reads out the semantic content specified by the semantic content ID “M00001” associated with the identification ID “0001” from the voice recognition dictionary 18. Thereby, the meaning content represented by the audio data is recognized.

  Specifically, when the user inputs a voice “Genzaichi” to the microphone 17, there are various voice data evaluation values (voice print evaluation value, accent evaluation value, pitch evaluation value, and utterance content evaluation value). The identification ID “0001” is hit by collating the voice ID parameter Pa, the accent parameter Pb, the pitch parameter Pc, and the utterance content parameter Pd of the unique identification ID and evaluating the similarity.

  Thereby, the voice recognition unit 24 determines that the meaning content of the input voice data is represented by the meaning content ID “M00001”, and displays the word “current location” on the touch panel 12. In this way, the voice recognition unit 24 recognizes that the voice data represents the meaning content “current location”.

  However, at least one of voiceprint similarity, accent similarity, pitch similarity, or speech content similarity satisfies voiceprint threshold Ta, accent threshold Tb, pitch threshold Tc, or speech content threshold Td. When there is no word, it is assumed that words similar to the voice data are not found in the voice recognition dictionary 18 and are input to the touch panel 12 or the remote control device 20 by the user. As described above, when an input is made to the touch panel 12 or the remote control device 20, the speech recognition learning unit 25 performs learning as follows.

  The voice recognition learning unit 25 receives a voice signal from the input analysis control unit 22 when a signal indicating that there is an input to the touch panel 12 or the remote control device 20 within a predetermined time after the voice is input to the microphone 17. The recognition method is learned based on the meaning content recognized by the recognition unit 24 and the content of the command input to the touch panel 12 or the remote operation device 20.

  A signal indicating that there is an input to the touch panel 12 or the remote control device 20 within a predetermined time (within 5 seconds) after the voice is input to the microphone 17 is input from the input analysis control unit 22 to the voice recognition learning unit 25. In this case, it is assumed that the touch panel 12 or the remote control device 20 is operated because voice recognition is not correctly performed.

  For this reason, the navigation device 10 according to the present embodiment learns the features of the user's voice to improve the accuracy of the voice recognition, and the meaning content recognized by the voice recognition unit 24 and the touch panel 12 or the remote Based on the similarity to the content of the command input to the controller device 20, the value of the voiceprint threshold Ta, the accent threshold Tb, the pitch threshold Tc, or the speech content threshold Td is changed in accordance with the user's voice characteristics.

  In this way, by changing one of the thresholds according to the characteristics of the user's voice, the evaluation criteria in the voice recognition process are changed, so that the user's utterance that was not correctly recognized before the change is correct. Voice recognition is started, and the accuracy of voice recognition can be improved.

  Here, when the user inputs the voice “Genzaichi” to the microphone 17, it is not determined that the evaluation value indicating the accent of the voice data is similar to the accent parameter Pb of the word “current location”. A case where voice recognition is not performed correctly will be described as a specific example.

  In this case, when the “current location” button on the touch panel 12 is pressed by the user within 5 seconds after the voice is input, the voice recognition learning unit 25 specifies the voiceprint threshold Ta, which specifies the word represented by the command input to the touch panel 12. The values of the accent threshold value Tb, the pitch threshold value Tc, and the speech content threshold value Td are read from the speech recognition dictionary 18.

  The speech recognition learning unit 25 compares the read threshold value (Ta to Td) with the similarity, and specifies which similarity is less than the threshold value.

  In this case, it is determined that the similarity of the voiceprint, the similarity of the pitch, and the similarity of the utterance content are not less than the values of the voiceprint threshold Ta, the pitch threshold Tc, and the utterance content threshold Td. Since the similarity is less than the accent threshold value Tb, the value of the accent threshold value Tb is changed.

  Here, if the accent threshold value Tb after change (after change), the accent threshold value Tb before change (before change), and the difference ΔTb between the accent threshold value Tb before change and the similarity between accents are used, Tb (after change) Is expressed as:

Tb (after change) = Tb (before change) −ΔTb
Here, ΔTb = Tb (before change) −K, and K is “accent similarity when speech recognition is not performed correctly”.

  That is, ΔTb is subtracted from Tb (before change). Thereby, the accent threshold value Tb (after change) is set to the same value as “accent similarity (K) when speech recognition is not performed correctly”. This is learning of a speech recognition method.

  Thereby, when the user inputs the voice “Genzaichi” to the microphone 17 next time, the voiceprint evaluation value, the accent evaluation value, the pitch evaluation value, the utterance content evaluation value, the voiceprint parameter Pa, and the accent Since it is determined that the parameter Pb, the pitch parameter Pc, and the speech content parameter Pd are similar, the speech recognition learning unit 25 correctly recognizes that the meaning content of the user's speech data is “current location”. Become.

  This is the same even when it is determined that any of the voiceprint evaluation value, the pitch evaluation value, or the speech content evaluation value is not similar to the voiceprint parameter Pa, the pitch parameter Pc, or the speech content parameter Pd. When the voice recognition threshold value Ta, the accent threshold value Tb, the pitch threshold value Tc, or the utterance content threshold value Td is changed by the voice recognition learning unit 25, voice recognition is correctly performed from the next time.

  FIG. 4 is a diagram showing a processing procedure of a learning process of a voice recognition method in the in-vehicle information system of the present embodiment. Note that the process shown in FIG. 2 is executed by the navigation ECU 11.

  When the power of the navigation device 10 is turned on, the navigation ECU 11 starts the processing procedure of the learning process of the recognition method in the in-vehicle information system of the present embodiment (start).

  The navigation ECU 11 converts the voice input to the microphone 17 into voice data (step S1). This process is a process executed by the function of the navigation ECU 11 as the input analysis control unit 22, and audio data is obtained by digitally converting the audio signal output from the microphone 17.

  The navigation ECU 11 performs voice recognition processing on the voice data (step S2). This voice recognition process is a process executed by the function of the navigation ECU 11 as the voice recognition unit 24, and recognizes the meaning content of the voice data transmitted from the input analysis control unit 22 using the voice recognition dictionary 18.

  Specifically, the voice data evaluation values (voice print evaluation value, accent evaluation value, pitch evaluation value, and speech content evaluation value) are calculated.

  Next, the navigation ECU 11 analyzes the evaluation result of the voice feature (step S3). This process is a process executed by the function of the navigation ECU 11 as the voice recognition unit 24. Specifically, as a result of the comparison in step S2, an evaluation value of voice data (voice print evaluation value, accent evaluation value, pitch evaluation). Data and voice content parameter Pa, and voice pattern parameter Pa, accent parameter Pb, pitch parameter Pc, and utterance content parameter Pd are all similar to each other. Is found, the semantic content word corresponding to the semantic content ID associated with the identification ID of the data is extracted from the speech recognition dictionary 18.

  Next, the navigation ECU 11 displays the word extracted in step S3 on the touch panel 12 (step S4). This is for user confirmation.

  Furthermore, navigation ECU11 monitors the input of the manual operation to the touch panel 12 or the remote control device 20 (step S5). This is because when voice recognition is not performed correctly, the user is likely to perform manual operation on the touch panel 12 or the remote control device 20.

  Next, the navigation ECU 11 determines whether or not a manual operation has been input to the touch panel 12 or the remote operation device 20 within 5 seconds after the sound is input to the microphone 17 (step S6). This is because if speech recognition is not performed correctly, learning of a recognition method is required.

  When the navigation ECU 11 determines that a manual operation has been input to the touch panel 12 or the remote control device 20 within 5 seconds after the voice is input to the microphone 17, the navigation ECU 11 analyzes the manual operation (step S7). For example, when the “current location” button displayed on the touch panel 12 is pressed, the navigation ECU 11 analyzes that the “current location” button is pressed.

  The navigation ECU 11 reads a threshold (Ta to Td) necessary for specifying the word represented by the operation content analyzed in step S7 from the speech recognition dictionary 18, and compares it with the similarity calculated for the speech data acquired in step S1. By doing so, it is specified whether the similarity that is less than the threshold (Ta to Td) is the similarity of the voiceprint, accent, pitch, or utterance content (step S8).

  The navigation ECU 11 subtracts ΔTb from the accent threshold Tb (before change) determined that the similarity is less than the threshold in step S8 (step S9). As a result, the value of the accent threshold value Tb becomes the same value as the accent similarity (K) specified in step S8. Here, ΔTb is a difference between Tb (before change) and “Accent similarity (K) when speech recognition is not correctly performed”.

  This is because, for example, when a voice “Genzaichi” is input to the microphone 17, it is determined that the evaluation value indicating the accent of the voice data is not similar to the accent parameter Pb of the word “current location”. This is a process of changing the value of the accent threshold value Tb so that it is determined that it is similar at the next speech recognition when it is not correctly performed.

  The navigation ECU 11 registers the accent threshold Tb (after change) changed in step S9 in the speech recognition dictionary 18 (step S10).

  Thereby, when the user inputs the voice “Genzaichi” to the microphone 17 next time, the voiceprint evaluation value, the accent evaluation value, the pitch evaluation value, the utterance content evaluation value, the voiceprint parameter Pa, and the accent Since it is determined that the parameter Pb, the pitch parameter Pc, and the speech content parameter Pd are similar to each other, the speech recognition learning unit 25 correctly recognizes that the meaning content of the user's speech data is “current location”. become.

  If it is determined in step S6 that no manual operation is input to the touch panel 12 or the remote control device 20 within 5 seconds after the sound is input to the microphone 17, the navigation ECU 11 determines in step S3 that the speech recognition dictionary The process of the navigation device 10 is executed using the command specified by the semantic content extracted from 18 (step S11).

  In this case, since the phrase having the meaning of “current location” is correctly recognized by voice, the navigation ECU 11 transmits a command specified by the phrase to a predetermined function unit in the navigation ECU 11.

  As described above, according to the navigation device 10 of the present embodiment, in order to improve the accuracy of speech recognition by learning the features of the user's speech, the meaning content recognized by the speech recognition unit 24, and the touch panel 12 or the voice print threshold value Ta, accent threshold value Tb, pitch threshold value Tc, or utterance content threshold value Td is changed in accordance with the characteristics of the user's voice based on the consistency with the content of the command input to the remote control device 12 or the remote control device 20. Therefore, the user's utterance that was not correctly recognized before the change is recognized correctly, and the accuracy of the speech recognition can be improved.

  In the above, the form in which the navigation ECU 11 performs the learning process of the speech recognition method has been described. However, the control device for realizing the in-vehicle information system of the present embodiment is realized by a dedicated ECU different from the navigation ECU. May be. This dedicated ECU may be disposed either inside or outside the navigation device 10.

  In the above description, the case where the touch panel 12 in which the liquid crystal monitor and the matrix switch are combined is used has been described. However, the touch panel type display unit configured to be operable when the input switch displayed on the panel is pressed by the user. If it is, it is not restricted to the touch panel 12 as mentioned above.

  In the above description, the in-vehicle information system is the navigation apparatus 10 and the learning function is added to the voice recognition function of the navigation apparatus 10. However, the in-vehicle information system is not limited to the navigation apparatus 10. When various in-vehicle devices such as audio and air conditioners have a voice recognition function, the in-vehicle information system of the present embodiment can be applied to the in-vehicle devices.

  As mentioned above, although the vehicle-mounted information system of exemplary embodiment of this invention was demonstrated, this invention is not limited to embodiment disclosed specifically, and does not deviate from a claim. Various modifications and changes are possible.

It is a figure which shows the structure of the vehicle-mounted information system of this Embodiment. It is a block diagram which shows the learning function of the speech recognition method in the vehicle-mounted information system of this Embodiment. It is a figure which shows the data structure of the speech recognition dictionary 18 in the vehicle-mounted information system of this Embodiment. It is a figure which shows the process sequence of the learning process of the speech recognition method in the vehicle-mounted information system of this Embodiment.

Explanation of symbols

10 Navigation Device 11 Navigation ECU
DESCRIPTION OF SYMBOLS 12 Touch panel 13 Current position detection part 14 Direction detection part 15 Route search part 16 Map database 17 Microphone 18 Voice recognition dictionary 19 Reception part 20 Remote operation device 21 Force management part 22 Input analysis control part 23 Manual input analysis part 24 Voice recognition part 25 Speech recognition learning unit

Claims (6)

First input means for inputting a command by manual operation;
A second input means for inputting voice;
Speech recognition means for recognizing speech input to the second input means;
Information processing means for outputting predetermined information based on the content of the command input to the first input means or the meaning content of the voice recognized by the voice recognition means;
When a command is input to the first input unit within a predetermined time after the voice is input to the second input unit, the data representing the content of the command and the voice input to the second input unit are An in-vehicle information system comprising learning means for learning a recognition method in the voice recognition means based on voice data to be expressed.   The learning means learns the characteristics of the voice input to the second input means based on the degree of coincidence between the data representing the content of the command input to the first input means and the voice data, The in-vehicle information system according to claim 1, wherein a recognition method in the voice recognition unit is learned. A speech recognition dictionary that stores a reference value for evaluating the characteristics of the speech and a threshold value for specifying the semantic content of the speech data;
The voice recognition unit compares the similarity between the evaluation value of the voice feature input to the second input unit and the reference value stored in the voice recognition dictionary with the threshold value, thereby obtaining the semantic content of the voice. Configured to recognize,
The learning means learns a recognition method in the voice recognition means by changing the threshold based on the degree of coincidence between the data representing the content of the command input to the first input means and the voice data. The in-vehicle information system according to claim 2.   The in-vehicle information system according to claim 2 or 3, wherein the feature of the voice is at least one of a voice print, an accent, a pitch, and an utterance content.   The in-vehicle information system according to any one of claims 1 to 4, wherein the first input means is a touch panel display means or a remote control device.   The in-vehicle information system according to any one of claims 1 to 5, wherein the information processing unit is included in an arithmetic processing unit of a navigation device.

Images