Summary of the invention
Technical matters to be solved by this invention is: the shortcoming at above prior art exists provides a kind of voice command discrimination high computing machine, and can facilitate for the user of beginner and physical handicaps.
The technical scheme that the present invention solves above technical matters is:
A kind of method of computer speech identification the steps include:
The I voice messaging is imported from voice-input device;
The II voice-input device carries out noise filtering to limiter and bandpass filter that voice messaging is sent to audio processing modules, the line data format conversion of going forward side by side;
The III voice messaging sends in the impact damper by audio processing modules;
The detection module of IV speech recognition system detects the voice messaging input in the audio processing modules, detection module sends instruction to analyzing identification module;
The analysis identification module of V speech recognition system calls the voice messaging and the memory instructions character library information of impact damper respectively and searches for comparison; As not finding match information in the instruction character library, then speech recognition system is in waiting status; As instruct the voice messaging that finds coupling in the character library, then enter the VI step; Instruction number in the described instruction character library is quantitative;
The instruction execution module of VI speech recognition system is opened corresponding program by the voice messaging of coupling, and CPU moves this program.
Voice messaging helps improving discrimination by noise filtering; Speech recognition system only need be discerned key word in the input information (i.e. instruction), and does not need to discern all input voice messagings, helps improving the identification accurate rate; If the quantity of instruction is unlimited, then roughly the same the quantity of instruction is far longer than quantitatively, instructs roughly the same that multicomputer is easy more more makes mistakes, and instruction number is that quantitative helping improves identification accuracy; Only need be by the said method user to the direct oral order of computing machine, computing machine just can be carried out by the voice instruction, provides convenience to the user like this, lays a good foundation for computing machine develops to the home entertaining direction simultaneously.
Technical scheme as the further qualification of the present invention:
Described analysis identification module call before the voice messaging of impact damper and the memory instructions character library information or described instruction execution module by the voice messaging of coupling open before the corresponding program (be step V or step VI before) further comprising the steps of: analyze voice messaging and the storer identifier word library information that identification module calls impact damper respectively and search for comparison, as not finding match information in the sign character library, then speech recognition system is in waiting status; As identify the voice messaging that finds coupling in the character library, then enter next step.Can avoid unconscious oral account to cause computerized information to be misread and false triggering like this, help improving identification accuracy.
The instruction execution module is opened before the corresponding program (be step VI before) by the voice messaging of coupling and is also comprised following three steps:
I instruction execution module is differentiated this instruction, is not special instruction as instruction, then directly opens the corresponding program that is identified instruction, and CPU moves this program simultaneously; For special instruction, then instruct execution module from hard disk is read the storehouse really, to call suggestion voice information as instruction, simultaneously voice messaging is sent to audio processing modules;
The ii audio processing modules sends instantaneous speech power to after the voice messaging Data Format Transform, the instantaneous speech power pronunciation;
Iii user presses suggestion voice information indicating input validation voice messaging once more, and audio processing modules sends speech data to impact damper then again to confirming that voice messaging carries out noise filtering and Data Format Transform; Detection module detects simultaneously the voice messaging input in the audio processing modules, detection module sends instruction to analyzing identification module; Analyze identification module and receive that detection module instructs back read buffers information and storer affirmation character library information to go forward side by side line search relatively, as do not find the voice messaging of coupling, then speech recognition system is in waiting status; As find the voice messaging of coupling, and then instruct execution module to open corresponding program by this voice messaging, CPU moves this program.
The present invention solves the computing machine that above technical matters also provides a kind of energy voice command recognition, comprises speech input device, audio processing modules, impact damper, storer, CPU and speech recognition system; Audio processing modules comprises: A/D converter and D/A converter; Speech recognition system comprises: analyze identification module and instruction execution module;
Audio processing modules also comprises: limiter and bandpass filter;
Limiter is used to filter the noise that hangs down decibel and high-decibel, is positioned in the middle of speech input device and the bandpass filter;
Bandpass filter is used to filter low decibel and the above noise of high-decibel, between limiter and A/D converter;
Speech recognition system also comprises detection module; Detection module is used to detect the voice messaging input condition, and the notification analysis identification module is audio processing modules and analysis identification module connector simultaneously;
Storer comprises: the instruction character library;
The instruction character library is used for the voice messaging of storage instruction, and other voice messaging is not stored, and instruction number is quantitative, and instruction is to be deposited in advance by the equipment supplier, and the user can not revise; Instruction database carries out alternately with the analysis identification module;
Technical scheme as the further qualification of the present invention:
For fear of unconscious oral account, further improve the identification accurate rate simultaneously, find out key word in user's oral order statement, computing machine only needs identidication key, and not needing to discern all input informations, voice messaging finally is simplified to " instruction " or " identifier word ";
Storer also is provided with the sign character library; The sign character library is used for the storaging mark word, and the user sets up storage on their own, and identification module is mutual with analyzing;
Computing machine also comprises instantaneous speech power, and the computer-internal information command is exported by instantaneous speech power, is the external interface of computer-internal information output;
Storer also is provided with the affirmation character library; Confirm character library, be used for the storaged voice information and confirm word, and confirm word for reading really certainly, such as "Yes", " will " etc.; Carry out alternately with analysis identification module, instruction execution module;
When the instruction execution module identifies instruction is ordinary instruction, then do not need the user to reaffirm, input ordinary instruction computing machine can not influence execution speed, and needs the user to reaffirm the raising precision at special instruction, brings some unnecessary losses to avoid slip of the tongue to the user.
Embodiment
In order to understand essence of the present invention better, following examples are described in detail the present invention in conjunction with the accompanying drawings.
Embodiment one
The invention provides a kind of computing machine of energy voice command recognition, as shown in Figure 1, audio recognition method is installed realization on computers.It comprises speech input device, instantaneous speech power, audio processing modules, impact damper, storer, CPU, speech recognition system; Audio processing modules comprises A/D converter, D/A converter, limiter, bandpass filter; Speech recognition system comprises: detection module, analysis identification module, instruction execution module; Storer comprises: instruction character library, sign character library, affirmation character library;
Language input dress is the interface that user's voice information is input to computer-internal, and speech input device can be microphone, microphone etc.;
The limiter of audio processing modules falls the noise filtering below 40 decibels and more than 60 decibels, is positioned in the middle of speech input device and the bandpass filter, and voice messaging is transported to limiter by speech input device, is transported to bandpass filter by limiter again;
The bandpass filter of audio processing modules, front end being sneaked into AC power noise and the outer noise filtering of voice band of 50Hz falls, with free transmission range for being arranged on 150~6000HZ, between limiter and A/D converter, voice messaging is transported to bandpass filter by limiter, is transported to A/D converter by bandpass filter again;
The A/D converter of audio processing modules is used for the converting speech format information memory, and between bandpass filter and impact damper, voice messaging is transported to A/D converter by bandpass filter, is transported to impact damper by A/D converter again;
The D/A converter of audio processing modules is used for the converting speech format information memory, and suggestion voice information outputs to instantaneous speech power again from confirming that character library forwards D/A converter to;
Instantaneous speech power is the external interface of computer-internal information output, and the computer-internal information command is exported by instantaneous speech power, and instantaneous speech power can be loudspeaker, audio amplifier etc.;
Impact damper is used for the voice messaging after the stores processor, and voice messaging outputs to impact damper from A/D converter, and the analysis identification module with speech recognition system carries out alternately again;
The detection module of speech recognition system is used to detect audio processing modules and has or not the phonetic entry situation, and the notification analysis identification module is audio processing modules and analysis identification module connector simultaneously;
The analysis identification module of speech recognition system, be used to search for the voice messaging of comparison impact damper and the voice messaging and the identification in the corresponding storehouse of storer, analyze instruction character library, sign character library, the affirmation character library of identification module and storer and carry out respectively alternately, analyze identification module simultaneously and also carry out alternately with impact damper;
The instruction execution module of speech recognition system, whether be special instruction, be special instruction as instruction if being used for discrimination instruction, then instruct execution module to call the suggestion voice information that storer is confirmed character library, simultaneously voice output arrived audio processing modules; The instruction unpack corresponding program of instruction execution module by being identified, CPU moves this program;
The instruction character library of storer is used for the voice messaging of storage instruction, and other voice messaging is not stored, and is to be deposited in advance by the equipment supplier, and the user can not revise, and the quantity of storage instruction is quantitative; The sign character library of storer is used for the storaging mark word, and the user sets up storage on their own, can set according to user's hobby, such as " sesame ", " robot ", " meteor shower " etc.; This instruction character library, sign character library are mutual with the analysis identification module respectively;
The storer storehouse of reading really is used to store suggestion voice information and confirms word, and it all is sure word that storage is read really, such as "Yes", " will " etc., confirm character library and analyze identification module, the instruction execution module carries out alternately.
Example two
Fig. 2 illustrates a kind of process flow diagram of audio recognition method, the steps include:
The I voice messaging is imported from voice-input device;
The II voice-input device carries out noise filtering to limiter and bandpass filter that voice messaging is sent to audio processing modules, the line data format conversion of going forward side by side;
The III voice messaging sends in the impact damper by audio processing modules;
The detection module of IV speech recognition system detects the voice messaging input in the audio processing modules, detection module sends the analysis identification module of instruction to speech recognition system;
The analysis identification module of V speech recognition system calls the voice messaging and the memory instructions character library information of impact damper respectively and searches for comparison, and as not finding match information in the instruction character library, then speech recognition system is in waiting status; As instruct the voice messaging that finds coupling in the character library, then enter the VI step;
The instruction execution module of VI speech recognition system is opened corresponding program by the voice messaging of coupling, and CPU moves this program simultaneously.
For example the memory instructions character library at machine has " online " instruction, when user's input " I will surf the Net ", the analysis identification module of speech recognition system finds " online " instruction in the instruction character library, the instruction execution module is opened Internet routine, and CPU just moves Internet routine.
Example three
Fig. 3 illustrates the process flow diagram of another kind of audio recognition method, the steps include:
The I voice messaging is imported from voice-input device;
The II voice-input device carries out noise filtering to limiter and bandpass filter that voice messaging is sent to audio processing modules, the line data format conversion of going forward side by side;
The III voice messaging sends in the impact damper by audio processing modules;
The detection module of IV speech recognition system detects the voice messaging input in the audio processing modules, detection module sends the analysis identification module of instruction to speech recognition system;
The analysis identification module of V speech recognition system calls the voice messaging and the storer identifier word library information of impact damper respectively and searches for comparison; As not finding match information in the sign character library, then speech recognition system is in waiting status; As identify the voice messaging that finds coupling in the character library, then enter the VI step;
VI analysis identification module calls the voice messaging and the memory instructions character library information of impact damper respectively and searches for comparison; As not finding match information in the instruction character library, then speech recognition system is in waiting status; As instruct the voice messaging that finds coupling in the character library, then enter the VII step;
VII instruction execution module is differentiated this instruction, is not special instruction as instruction, then directly opens the corresponding program that is identified instruction, and CPU moves this program simultaneously; For special instruction, then instruct execution module from hard disk is read the storehouse really, to call suggestion voice information as instruction, simultaneously voice messaging is sent to audio processing modules;
The VIII audio processing modules sends instantaneous speech power to after the voice messaging Data Format Transform, the instantaneous speech power pronunciation;
IX user presses suggestion voice information indicating input validation voice messaging once more, and audio processing modules sends speech data to impact damper then again to confirming that voice messaging carries out noise filtering and Data Format Transform; Detection module detects simultaneously the voice messaging input in the audio processing modules, detection module sends the analysis identification module of instruction to speech recognition system; Analyze identification module and receive that detection module instructs back read buffers information and storer affirmation character library information to go forward side by side line search relatively, as do not find the voice messaging of coupling, then speech recognition system is in waiting status; As find the voice messaging of coupling, and then instruct execution module to open corresponding program by this voice messaging, CPU moves this program.
For example the user deposits the sign character library to " sesame " in, instruction word library storage " recovery system " instruction, in speech recognition system, special instruction is arranged in " recovery system " instruction simultaneously, is confirmed that library stores has voice messagings such as "Yes", " asking the phonetic entry yes or no "; When user speech input " sesame; I want recovery system ", call the voice messaging and the identifier word library information of impact damper by the analysis identification module and search for comparison, in sign after character library finds " sesame " information, analyze the voice messaging and instruction character library information that identification module begins to call impact damper and search for comparison; Find " recovery system " in the instruction character library, then instruct execution module to begin to differentiate according to the instruction voice messaging, final differentiation " recovery system " is a special instruction, execution module confirms that from storer character library calls suggestion voice information " please phonetic entry yes or no ", the D/A converter that suggestion voice information the is sent to audio processing modules row format conversion of going forward side by side; The instantaneous speech power sounding " is asked the phonetic entry yes or no ", user speech input "Yes", after handling, audio processing modules sends to impact damper, detection module detects audio processing modules has voice messaging input back to send instruction notification analysis identification module, at this moment analyze the voice messaging that identification module begins to call impact damper and search for comparison with the information of confirming character library, confirming that character library finds "Yes" information, the instruction execution module begins to open the recovery system program, and CPU is with regard to the restoring running system program.
The present invention is not limited to the foregoing description, with the identical or akin technical scheme of the technical characterictic of the foregoing description, also belongs to the protection domain of requirement of the present invention.