JP2005091611A - Information terminal, speech recognition server, and speech recognition system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce ineffective data communication by enhancing the reliability of speech data to be transmitted to a speech recognition server. <P>SOLUTION: In an onboard speech recognition system, which performs a speech recognition processing at a server side, an onboard information apparatus 200 decides an amount of noise of speech data based on the SN ratio of inputted speech data in a speech data determination part 207, and a determination part 203 decides whether the correct speech recognition of the speech data is possible or not based on information of noise occurrence factors such as the traveling speed of a vehicle, an opening/closing condition of a window, then only speech data, which is decided that the speech recognition is possible, to the speech recognition server 210 through a communication part 210. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an information terminal, a voice recognition server, and a voice recognition system using them.

  Information systems that use voice as input and output are used in various fields such as in-vehicle information systems and home network systems. A system that uses voice as input / output includes a mechanism for removing a noise signal included in input voice data and correctly recognizing the voice data.

  For example, the conventional in-vehicle voice recognition device described in Patent Document 1 removes a noise signal obtained through a noise collecting microphone from a voice signal obtained through a speaker microphone. Thus, a high S / N ratio is obtained and the speech recognition rate is increased.

  In addition, the conventional voice interactive navigation system described in Patent Document 2 performs a call between an in-vehicle terminal and a voice interactive server, and performs voice recognition processing on the server side.

Japanese Patent Laid-Open No. 2-77799 JP 2002-318132 A

  In the on-vehicle speech recognition apparatus described in Patent Document 1, it is difficult to completely remove noise because there are various causes of noise generation inside and outside the vehicle.

  Further, in the voice interactive navigation system described in Patent Document 2, it takes time until the recognition result is obtained due to the communication time with the server and the recognition processing time. Since an input operation and data retransmission are required, there is a problem that wasteful communication occurs, operability is deteriorated, and a communication fee is increased.

The present invention has been made to solve the above-described problems. In the case where voice recognition processing is performed on the server side, it is possible to increase the reliability of voice data transmitted to the server and reduce invalid data communication. The purpose is to obtain possible information terminals.
It is another object of the present invention to provide a voice recognition server that can reduce useless data communication with an information terminal.
It is another object of the present invention to obtain a voice recognition system using the information terminal and the voice recognition server as described above.

  The information terminal according to the present invention includes a determination unit that determines whether or not the voice data input via the voice input unit is correctly recognized, and only the voice data that is determined to be correctly recognized by the determination unit. Is provided with a communication unit that transmits the message to the voice recognition server.

  A speech recognition server according to the present invention includes a speech recognition unit that performs speech recognition of speech data received from an information terminal, and a recognition result determination that compares a recognition result by the speech recognition unit with dictionary data and outputs a reliability of the recognition result. If the reliability is equal to or higher than a certain threshold, response data having response contents for the received voice data is created. If the reliability does not reach the threshold, whether the received voice data is recognized correctly A response data creating unit that creates response data having contents to be confirmed and a communication unit that transmits the response data created by the response data creating unit to the information terminal are provided.

  In the speech recognition system according to the present invention, the information terminal determines whether the speech data input via the speech input unit is correctly recognized by the speech recognition server in the determination unit, and is correctly recognized by the determination unit. Only the voice data determined to be transmitted to the voice recognition server via the communication unit, and the voice recognition server performs voice recognition of the voice data received from the information terminal via the communication unit and recognizes In the result determination unit, the recognition result by the speech recognition unit is compared with dictionary data to output the reliability of the recognition result. If the reliability is greater than a certain threshold, the response data creation unit receives the received speech data Create response data with the response contents for, and if the reliability does not reach the threshold, check whether the recognition result of the received voice data is correct Creates a response data capacity, and transmits via the communication unit the response data to the information terminal.

  According to the present invention, in the information terminal, the noise amount of the input voice data itself is determined, and the recognizability of the voice data is determined based on the information of the elements that cause various noises. Since only the voice data determined to be transmitted to the voice recognition server is transmitted, it is possible to eliminate as much as possible voice recognition failure in the voice recognition server and reduce useless communication.

  According to the present invention, when the voice recognition server determines that the voice recognition is successful, the response data including the response content to the transmission data is transmitted to the information terminal, and when it is determined that the voice recognition has failed, the information Since the response data having the content for confirming the content of the transmission data is transmitted to the terminal, it is possible to avoid unnecessary data transmission from the voice recognition server.

  According to the present invention, the information terminal determines the recognizability of the sound data based on the noise amount of the input sound data itself and the information of the elements that cause various noises, and the sound that is determined to be recognizable When only the data is transmitted to the voice recognition server and the voice recognition server determines that the voice recognition of the received voice data is successful, the voice recognition server transmits response data including a response content to the transmission data to the information terminal. When it is determined that it has failed, response data with the content to confirm the content of the transmission data is transmitted to the information terminal, so that it is possible to reduce invalid data communication and improve user operation efficiency Voice recognition system can be obtained.

Hereinafter, various embodiments of the present invention will be described.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of an in-vehicle speech recognition system 100 according to Embodiment 1 of the present invention. The voice recognition system 100 includes an in-vehicle information device (information terminal) 200 and a voice recognition server 300. The in-vehicle information device 200 and the voice recognition server 300 are connected via a wireless communication line.

  The in-vehicle information device 200 includes a vehicle information acquisition unit 201, a noise amount determination unit 202, a determination unit (noise amount determination unit, determination unit) 203, a threshold storage unit 204, an audio input unit 205, an S / N ratio acquisition unit 206, an audio A data determination unit 207, a voice output unit 208, a control unit 209, a communication unit 210, a recognition result determination unit 211, and a determination condition learning unit 212 are provided. The in-vehicle information device 200 is connected to the microphone 11 of the in-vehicle information terminal via the voice input unit 205 and the speaker 10 via the voice output unit 208. In addition, the vehicle information acquisition unit 201 is connected to the vehicle speedometer 12, the window opening / closing device 13, the wiper driving device 14, the winker 15, the car navigation device 16, the in-vehicle audio device 17, and the in-vehicle air conditioner 18. The control unit 209 controls the entire in-vehicle information device 200.

  Vehicle information acquisition unit 201, noise amount determination unit 202, determination unit 203, threshold storage unit 204, audio input unit 205, S / N ratio acquisition unit 206, audio data determination unit 207, audio output unit 208, control unit 209, communication The unit 210, the recognition result determination unit 211, and the determination condition learning unit 212 constitute a part of the central processing unit of the in-vehicle information device 200, and these operate the central processing unit of the in-vehicle information device 200. It corresponds to the program module to be controlled.

  The speech recognition server 300 includes a communication unit 301, a control unit 302, a speech recognition unit 303, a recognition result determination unit 304, and a response data creation unit 305. In addition, a recognition dictionary storage unit 306 is connected to the voice recognition server 300. The recognition dictionary storage unit 306 is a storage device that stores a recognition dictionary database. Further, the control unit 302 controls the entire voice recognition server 300.

  The communication unit 301, the control unit 302, the speech recognition unit 303, the recognition result determination unit 304, and the response data creation unit 305 constitute part of the central processing unit of the speech recognition server 300. This corresponds to a program module that controls the operation of the central processing unit of the server 300.

Next, the operation will be described.
FIG. 2 is a flowchart of voice recognition processing by the in-vehicle information device 200 according to the first embodiment.
When the voice input process is started, the voice spoken by the user is acquired via the microphone 11 (step ST101), and voice data is input to the voice input unit 205 (step ST102).

  Next, the S / N ratio acquisition unit 206 acquires the S / N ratio (N) of the input audio data (step ST103). The S / N ratio (N) is calculated by comparing the ambient noise signal acquired by the in-vehicle information device 200 immediately before the start of the voice input process with the voice signal input in step ST102.

  Next, the audio data determination unit 207 compares the S / N ratio (N) acquired in step ST103 with the threshold (N0) (step ST104). If it is determined in step ST104 that the S / N ratio (N) is greater than or equal to the threshold (N0), it is determined that the input audio data is noisy and misrecognition is likely to occur. Prompt the user to re-enter.

When it is determined in step ST104 that the S / N ratio (N) is smaller than the threshold value (N0), the vehicle information acquisition unit 201 acquires information on various elements that cause noise in the vehicle. (Step ST105).
For example, the traveling speed is acquired from the vehicle speedometer 12. The window opening / closing device 13 acquires the opening / closing state of the window. As for the open / closed state of the window, information is acquired indicating whether the window is open or closed, and if it is open, how much is open. Further, the wiper driving device 14 acquires information about whether the wiper is moving or stopped, and how fast the wiper is moving when it is moving. Further, from the winker 15, information on whether the winker is operating or stopped is acquired. From the car navigation device 16, information on the current travel point, that is, the travel point at the time of voice data input is acquired. As information on the travel point, for example, information on whether there is a construction site in the immediate vicinity, such as the state of the road on which the vehicle is traveling, that is, whether or not the vehicle is paved is acquired. If it is in use, the set volume is acquired from the in-vehicle audio device 17. If the air conditioner is in operation, the set air flow rate is acquired from the in-vehicle air conditioner 18.

  The noise amount determination unit 202 acquires a noise amount determination value S for each element based on information about various elements that cause noise generation acquired by the vehicle information acquisition unit 201 (step ST105). As the determination value S, the traveling speed acquired by the noise amount determination unit 202 may be used as it is, or the open / close state of the window, the operation state of the wiper, and the like may be scored in a plurality of stages. For example, the open / closed state of the window may be set to 0 point if it is closed, 1 point if the open part is less than half, 2 points if it is more than half open, and 3 points if it is fully open. . Similarly, the wiper may be set to 0 point if it is stopped, 1 point if it is moving at the slowest setting speed, 2 points if it is an intermediate speed, and 3 points if it is the fastest speed.

  Next, the determination unit 203 acquires the noise amount determination value S obtained in step ST105 and also acquires the noise amount threshold value S0 for each element from the threshold storage unit 204 (step ST106).

FIG. 3 is a diagram for explaining a method of setting the noise amount threshold value S0 stored in the threshold value storage unit 204. Here, a description will be given by taking the traveling speed of the vehicle as an example of an element that causes noise in the vehicle.
The threshold value storage unit 204 holds a threshold value S0 obtained according to a predetermined recognition rate or a noise amount limit value. In the example shown in the figure, whether or not the speech recognition rate is 50% or more is defined as a limit value A for determining whether or not to transmit to the speech recognition server 300. In this case, the speech recognition rate is 80 km / h, which is 50%. Becomes the threshold S0 of the traveling speed.
In addition to the traveling speed, the limit value A is similarly applied to other factors such as the opening / closing state of the window, the acceleration state, the audio volume state, the wiper operation state, the blinker operation state, the air conditioner air flow state, and the travel point state. A value satisfying the condition is defined as a threshold value S0.

  Further, when there are a plurality of elements that cause noise generation, a more appropriate threshold value S0 can be obtained by setting a threshold value S0 that satisfies the recognition rate or the noise amount limit value in a state in which these conditions are combined. be able to. FIG. 4 shows an example of a method for setting the threshold value S0 of the traveling speed when the traveling speed and the window open / closed state are combined. As shown in the figure, the running speed threshold S0 is 80 km / h when the window is closed, but the running speed that satisfies the recognition rate limit value of 50% when the window is half open and further fully opened. The threshold value S0 is 70 km / h and 65 km / h.

The determination unit 203 compares the determination value S with the threshold value S0 for each element that causes noise generation (step ST107).
If it is determined in step ST107 that the determination value S is greater than or equal to the threshold value S0, it is determined that the input voice data has a large amount of noise and misrecognition is likely to occur, and the voice data is transmitted to the voice recognition server 300. Without prompting the user to input again. If it demonstrates using the example of FIG. 3, when the driving speed of the vehicle acquired by step ST105 is 80 km / h or more, audio | voice data will not be transmitted to the speech recognition server 300, but a user will be prompted for re-input.

  If it is determined in step ST107 that the determination value S is smaller than the threshold value S0, the communication unit 210 transmits the sound data to the sound recognition server 300 (step ST108). If it demonstrates using the example of FIG. 3, if the driving speed of the vehicle acquired by step ST105 is 60 km / h, since it is smaller than the threshold value 80 km / h, audio | voice data will be transmitted to the speech recognition server 300. FIG.

  When the voice recognition process described later is performed in the voice recognition server 300 and response data is notified from the voice recognition server 300, the communication unit 210 receives the response data (step ST109).

The recognition result determination unit 211 determines the content of the received response data, and determines whether or not the voice recognition is correctly performed in the voice recognition server 300 (step ST110).
If it is determined in step ST110 that the speech recognition by the speech recognition server 300 is successful, the determination condition learning unit 212 increases the value of the threshold value S0 stored in the threshold value storage unit 204 (step ST111). On the other hand, when it is determined in step ST110 that the voice recognition has failed, the threshold value S0 is decreased (step ST112).

  The determination condition learning unit 212 updates the threshold value S0 for each vehicle state held in the threshold value storage unit 204 based on whether or not the voice recognition by the voice recognition server 300 is successful. For example, in the example of FIG. 3, the threshold value S0 of the traveling speed is set to 80 km / h, but when the voice recognition by the voice recognition server 300 fails, the threshold value S0 is lowered to 60 km / h. As a result, the condition for determining whether or not to transmit the voice data to the voice recognition server 300 becomes strict, and only the voice data with higher reliability can be transmitted to the voice recognition server 300.

Next, the operation of the voice recognition server 300 will be described.
FIG. 5 is a flowchart of voice recognition processing by the voice recognition server 300 according to the first embodiment.
The communication unit 301 receives the audio data transmitted by the in-vehicle information device 200 in step ST108 (step ST201).

  Next, the speech recognition unit 303 performs speech recognition processing on the received speech data based on the dictionary data stored in the recognition dictionary storage unit 306 (step ST202).

  Next, the recognition result determination unit 304 uses a result score s (reliability) indicating how much the result of the speech recognition performed in step ST202 matches the dictionary data stored in the recognition dictionary storage unit 306. Output (step ST203).

Next, the score s acquired in step ST203 is compared with a preset score threshold s0 (step ST204).
If it is determined in step ST204 that the score s is greater than or equal to the threshold value s0, it is determined that there is a high possibility of misrecognition, and the response data creation unit 305 determines whether the voice recognition result is correct or not. Response data having contents to be confirmed is created (step ST206).
On the other hand, if it is determined in step ST204 that the score s is smaller than the threshold value s0, it is determined that there is a high possibility that the speech recognition is successful, and the response data including the response content for the recognition based on the recognition result. Is created (step ST205).

  Next, the communication unit 301 transmits the response data created in step ST205 or step ST206 to the in-vehicle information device 200 (step ST207). The transmitted response data is received by the in-vehicle information device 200 in step ST109. In step ST110, the recognition result determination unit 211 of the in-vehicle information device 200 determines whether or not the speech recognition by the speech recognition server 300 has succeeded or failed based on the content of the received response data.

  The flow of voice recognition processing between the in-vehicle information device 200 and the voice recognition server 300 according to the first embodiment will be described with reference to FIGS. 6 and 7. FIG. 6 shows a sequence in the case where the speech recognition server 300 determines that the speech recognition result score s is smaller than the threshold value s0 in step ST204, that is, it is determined that the speech recognition is successful. For example, when the in-vehicle information device 200 transmits voice data having the content “near ramen shop” to the voice recognition server 300, if the voice recognition server 300 determines that the score s of the recognition result is smaller than the threshold value s0, the voice In response to the transmitted voice data request, the recognition server 300 searches for a ramen shop, and transmits the search result ramen shop list to the in-vehicle information device 200 as response data. The audio output unit 208 of the in-vehicle information device 200 outputs response data received via the speaker 10. At this time, since the voice recognition by the voice recognition server 300 is successful, the determination condition learning unit 212 of the in-vehicle information device 200 increases the value of the threshold value S0 stored in the threshold value storage unit 204.

  On the other hand, FIG. 7 shows a sequence when the speech recognition server 300 determines that the speech recognition result score s is greater than or equal to the threshold value s0 in step ST204, that is, it is determined that speech recognition has failed. In this case, since there is a possibility that the voice recognition server 300 has erroneously recognized the content of the voice data, the response data creation unit 305 responds to the voice “near ramen shop” of the in-vehicle information device 200. For example, response data for confirmation such as “Is it a nearby ramen shop?” Is created and transmitted to the in-vehicle information device 200. The audio output unit 208 of the in-vehicle information device 200 outputs response data received via the speaker 10. For example, when the user answers “Yes” to the output voice, the voice data having the content “Yes” is transmitted to the voice recognition server 300. The voice recognition server 300 confirms the voice data with the content “Yes”, and transmits the ramen shop list of the search result to the in-vehicle information device 200. As a result, the communication procedure until the user obtains the target data can be minimized, and unnecessary communication due to erroneous recognition can be avoided.

  As described above, according to the first embodiment, in the in-vehicle information device 200, the audio data determination unit 207 determines the noise amount of the input audio data itself, and the determination unit 203 causes various noise generation factors. Overall, the possibility of recognizing the voice data is determined and the transmission to the voice recognition server 300 is controlled. Therefore, the voice recognition failure in the voice recognition server 300 can be eliminated as much as possible. For this reason, useless communication can be reduced and communication charges can be reduced. In addition, useless user operations due to occurrence of erroneous recognition can be omitted, and the operation efficiency is improved.

  When the voice recognition server 300 determines that the voice recognition is successful, the voice recognition server 300 transmits response data including the response content to the transmission data to the in-vehicle information device 200. When the voice recognition server 300 determines that the voice recognition has failed, the in-vehicle information Since the response data having the content for confirming the content of the transmission data is transmitted to the device 200, it is possible to avoid unnecessary data transmission from the voice recognition server 300.

  Furthermore, the in-vehicle information device 200 determines whether or not the speech recognition at the speech recognition server 300 is successful or unsuccessful by the recognition result determination unit 211 based on the response data received from the speech recognition server 300. Since the threshold value stored in the threshold value storage unit 204 is increased and the threshold value is decreased if the threshold value is unsuccessful, the reliability of the voice data to be transmitted to the voice recognition server 300 can be further increased, and more than necessary. It is possible to prevent the transmission conditions from becoming severe.

  In the first embodiment, the function of the car navigation device 16 is used to acquire travel point information and use it to determine the amount of noise. Therefore, the car navigation device 16 can be used effectively to achieve higher accuracy. A high voice recognition system can be realized.

Embodiment 2. FIG.
FIG. 8 is a block diagram showing a configuration of a voice recognition system 500 on the home network according to the second embodiment of the present invention. The voice recognition system 500 includes an information terminal 600 on a home network 900 and a voice recognition server 700. The information terminal 600 and the voice recognition server 700 are connected via a communication line. The configurations of the information terminal 600 and the voice recognition server 700 are the same as those of the in-vehicle information device 200 and the voice recognition server 300 of the first embodiment.

  The information terminal 600 is connected to the home server 800 of the home network 900. For example, a television 801, an audio device 802, a washing machine 803, an air conditioner 804, and a personal computer 805 are connected to the home server 800, and the home server 800 centrally manages the electrical devices and information devices in the house. .

Next, the operation of the speech recognition system 500 according to Embodiment 2 will be described.
The operations of information terminal 600 and voice recognition server 700 are the same as the operations shown in the flowcharts of FIGS. 2 and 5 of the first embodiment. In the second embodiment, information regarding an element that causes noise generation is acquired via the home server 800 by the same function as the vehicle information acquisition unit 201 of the in-vehicle information device 200. Specifically, the information includes the volume information of the television 801, the volume information of the audio device 802, the operating state of the washing machine 803, the air volume of the air conditioner 804, and the like. The noise amount determination unit 202 outputs each noise amount determination value S based on the acquired information. The determination unit 203 compares the noise amount determination value S with the threshold value S0 of each element held in the threshold value storage unit 204, and determines whether or not to transmit to the voice recognition server 700.

  As described above, according to the second embodiment, when information is exchanged with the voice recognition server 700 using the information terminal 600 in the house, the information terminal 600 can use the home server 800 in the house. Since the state information of various electrical devices that cause noise generation is collected, the amount of noise in the voice data is determined, and transmission to the voice recognition server 700 is controlled. Failure can be eliminated as much as possible. For this reason, useless communication can be reduced and communication charges can be reduced. In addition, useless user operations due to occurrence of erroneous recognition can be omitted, and the operation efficiency is improved.

  In the second embodiment, the client of the speech recognition system 500 is mounted on the information terminal 600, but may be mounted on an electrical device such as the television 801 or the personal computer 805 or the home server 800, for example.

It is a block diagram which shows the structure of the vehicle-mounted speech recognition system by Embodiment 1 of this invention. It is a flowchart of a process of the vehicle-mounted information apparatus by Embodiment 1 of this invention. It is a figure for demonstrating the noise amount determination conditions by Embodiment 1 of this invention. It is a figure for demonstrating the noise amount determination conditions by Embodiment 1 of this invention. It is a flowchart of the process of the speech recognition server by Embodiment 1 of this invention. It is a figure which shows the sequence of the speech recognition process by Embodiment 1 of this invention. It is a figure which shows the sequence of the speech recognition process by Embodiment 1 of this invention. It is a block diagram which shows the structure of the speech recognition system on a home network by Embodiment 2 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 Speaker, 11 Microphone, 12 Speedometer, 13 Window opening and closing device, 14 Wiper drive device, 15 Winker, 16 Car navigation device, 17 Car audio equipment, 18 Car air conditioner, 100,500 Voice recognition system, 200 Car information system (Information terminal), 201 vehicle information acquisition unit, 202 noise amount determination unit, 203 determination unit (noise amount determination unit, determination unit), 204 threshold storage unit, 205 voice input unit, 206 S / N ratio acquisition unit, 207 voice Data determination unit (determination unit), 208 voice output unit, 209 control unit, 210 communication unit, 211 recognition result determination unit, 212 determination condition learning unit, 300,700 speech recognition server, 301 communication unit, 302 control unit, 303 voice Recognition unit, 304 Recognition result determination unit, 305 Response data creation unit, 306 Recognition word Storage unit, 600 information terminal, 800 home server ,, 801 TV, 802 audio equipment, 803 washing machines, 804 Air Conditioning, 805 PC, 900 home network.

Claims (13)

A determination unit that determines whether or not the voice data input through the voice input unit is correctly recognized;
An information terminal including a communication unit that transmits only voice data that is determined to be correctly voice-recognized by the determination unit to a voice recognition server. The determination unit is a voice data determination unit that determines the reliability of the voice data itself based on the SN ratio between the input voice data and the ambient noise when the voice data is input;
The information terminal according to claim 1, further comprising a noise amount determination unit that determines whether correct voice recognition of the voice data is possible based on information about an element that causes noise generation.   The information terminal is an in-vehicle information device, and the noise amount determination unit determines whether correct voice recognition is possible based on information about a travel point at the time of voice data input acquired from a car navigation system. 2. Information terminal according to 2.   3. The information according to claim 2, wherein the information terminal is an in-vehicle information device, and the noise amount determination unit determines whether correct voice recognition is possible based on a traveling speed at the time of voice data input acquired from a speedometer. Terminal.   3. The information terminal according to claim 2, wherein the information terminal is an in-vehicle information device, and the noise amount determination unit determines whether correct voice recognition is possible based on an open / closed state of the vehicle window when voice data is input.   3. The information terminal according to claim 2, wherein the information terminal is an in-vehicle information device, and the noise amount determination unit determines whether correct voice recognition is possible based on a volume of the audio device at the time of inputting voice data.   3. The information terminal according to claim 2, wherein the information terminal is an in-vehicle information device, and the noise amount determination unit determines whether correct voice recognition is possible based on an operation state of the wiper when voice data is input.   3. The information terminal according to claim 2, wherein the information terminal is an in-vehicle information device, and the noise amount determination unit determines whether correct voice recognition is possible based on whether or not the winker at the time of voice data input is operating. Information terminal.   3. The information terminal according to claim 2, wherein the information terminal is an in-vehicle information device, and the noise amount determination unit determines whether or not correct voice recognition is possible based on an air volume of the in-vehicle air conditioner when voice data is input. .   The information terminal is an information terminal on a home network, and the noise amount determination unit determines whether correct voice recognition is possible based on an operating state of a device connected to the home network when voice data is input. The information terminal according to claim 2. A recognition result determination unit that determines whether the voice data transmitted to the voice recognition server is correctly voice-recognized by the voice recognition server based on the content of the response data from the voice recognition server;
When it is determined that the speech recognition is correctly recognized by the recognition result determination unit, the condition for determining that the speech data can be recognized by the noise amount determination unit is relaxed, and when it is determined that the speech recognition is not correctly performed, The information terminal according to any one of claims 2 to 10, further comprising a determination condition learning unit that makes conditions for determination stricter. A voice recognition unit that performs voice recognition of voice data received from the information terminal;
A recognition result determination unit that compares the recognition result by the voice recognition unit with dictionary data and outputs the reliability of the recognition result;
If the reliability is greater than or equal to a threshold value, response data having response contents to the received voice data is created. If the reliability does not reach the threshold value, the recognition result of the received voice data is A response data creation unit that creates response data with content to check whether it is correct,
A speech recognition server comprising a communication unit that transmits response data created by the response data creation unit to the information terminal. A speech recognition system comprising a speech recognition server and an information terminal,
In the information terminal, the determination unit determines whether or not the voice data input via the voice input unit is correctly recognized by the speech recognition server, and the sound determined to be correctly recognized by the determination unit Send only data to the voice recognition server via the communication unit,
The voice recognition server performs voice recognition of voice data received from the information terminal via a communication unit in a voice recognition unit, and compares a recognition result by the voice recognition unit with dictionary data in a recognition result determination unit. To output the reliability of the recognition result, and when the reliability is equal to or greater than a certain threshold, the response data generating unit generates response data having response contents for the received voice data, and the reliability is When the threshold value is not reached, response data having a content for confirming whether or not the recognition result of the received voice data is correct is created, and the response data is transmitted to the information terminal via a communication unit. Voice recognition system.

Images