CN102682767B - Speech recognition method applied to home network - Google Patents
Speech recognition method applied to home network Download PDFInfo
- Publication number
- CN102682767B CN102682767B CN201110065918.0A CN201110065918A CN102682767B CN 102682767 B CN102682767 B CN 102682767B CN 201110065918 A CN201110065918 A CN 201110065918A CN 102682767 B CN102682767 B CN 102682767B
- Authority
- CN
- China
- Prior art keywords
- stage
- voice
- recognition method
- speech
- recognition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Telephonic Communication Services (AREA)
Abstract
The invention relates to a speech recognition method used in home network and particularly relates to a method for performing speech recognition applied to home network through a speech recognizer. According to the invention, since the mode of the speech recognizer can be changed according to the peripheral situation, the error recognition to non-instruction speech is reduced by processing the speech recognition in time, therefore, the instruction recognition rate in the noise condition is enhanced, the tone of the speaker un-stored in a speed recognition database is adapted automatically, the acoustic model is recorded and recognized so as to enhance the recognition rate, the same instructions in a plurality of areas can be unified, the too many instructions are reduced, the accuracy of the speech recognition is greatly enhanced, and the error recognition is reduced effectively.
Description
Technical field
The present invention relates to a kind of audio recognition method, especially a kind of method of being carried out speech recognition by speech recognition machine.
Background technology
Along with the prosperity of network technology, the system using an input/output device to control the remote machine linked together by network is also gone on the market thereupon.Moreover, the system controlling the machine connected by network with phonetic order is also increasing.
Speech recognition system in conventional art exists noise affects defect and the limit such as apparatus control and the reduction of special sound person phonetic recognization rate.Reduce this point with regard to special sound person phonetic recognization rate, although this defect can be made up by adaptation, bring many inconvenience to user.Moreover, in order to control multiple machine by voice, need to store different instructions in each machine, thus cause instruction too much, too loaded down with trivial details, cause discrimination to reduce.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of audio recognition method being applied to home network, audio recognition method controls the machine connected by network by speech recognition, voice can be inputted at any time, prepare especially without the need to other, namely by other sound of speech recognition instant recognition instruction and non-instruction exactly, thus reduce the frequency of wrong identification, and automatically adapt to the voice of voice person; Moreover, the phonetic recognization rate of the person that can also improve special sound, the steering order that simultaneously can realize multiple machine unitizes, and improves the accuracy rate of speech recognition.
The technical scheme that the present invention solves the problems of the technologies described above is as follows: audio recognition method, comprising: the instant stage S1 receiving phonetic entry; Measure the power of the described voice inputted in the above-mentioned S1 stage and the stage S2 of duration; By the power of described voice that measures in the described S2 stage and duration compared with arbitrary value, and by the stage S3 that silent sound, off beat, forte are classified to described voice; According to the classification in the described S3 stage, if silent sound, then calculate the duration of silent sound, and continue to maintain phonetic entry holding state, if off beat or forte, then carry out the stage S4 of voice recognition processing; According to the voice recognition processing that the described S4 stage carries out, carry out the stage S5 formulating unit voice recognition processing; According to the result in described S5 stage, determine whether voice recognition processing is carried out to overall voice, the stage S6 that row relax of going forward side by side prepares; According to the described S6 stage, perform the voice recognition processing of overall voice if determine, then overall voice are carried out to the stage S7 of voice recognition processing; According to the result in described S7 stage, determine the result of whether certification voice recognition processing, and control machine, carry out the stage S8 identifying voice adaptation; In the described S6 stage, when speech recognition not being carried out to described overall voice, or according to the described S8 stage, during the recognition result of voice described in not certification, then according to the Classification of Speech determined in the described S3 stage, if during off beat, be then converted to noise pattern, if during forte, be then converted to the stage S9 of refusal pattern.
Further, be preferably also included in the described S3 stage, according to the power formulated in advance, described voice can be divided into Three Estate, if during the minimum the first estate of the speech volume measured in the described S2 stage, then silent sound class assigned in described voice; If the volume of described voice belongs to the second grade higher than described the first estate, and when the duration of a sound of described voice is arbitrary value, then off beat class assigned in described voice; If the volume of described voice is the highest tertiary gradient, and when the duration of a sound of described voice meets arbitrary value, then the S10 stage of forte class assigned in described voice.
Further, in the process that the voice recognition processing being preferably also included in the described formulation unit in described S5 stage is carried out, if during identidication key, the log-likelihood ratio and the respective acoustic model that described key word most end phoneme are transmitted to virtual machine contrast, and memory has the stage S11 of the acoustic model of mxm..
Preferably also be included in the described S6 stage, institute's speech recognition result with to deposit instruction consistent, and determines to carry out to described overall voice the stage S12 that second time identifies.
Preferably also be included in the described S12 stage, when determining to carry out described second time speech recognition, the described acoustic model remembered be changed to the preparatory stage S13 of second time recognition mode in the described S11 stage.
Preferably also be included in the described S13 stage, the described acoustic model prepared is used for database, and carry out the stage S14 of second time speech recognition.
Preferably also comprise: in the described S8 stage, if described second time voice identification result is certified, according to the stage S15 of described authentication result control machine; And in the described S8 stage, store voice, and the adaptation operation carrying out stored voice, upgrade the stage S16 with the acoustic model of the tamber characteristic of the voice person of pronunciation.
Further, preferably also comprise: in the described S16 stage, check whether the stage S17 of the database with described voice person tamber characteristic; And in the described S16 stage, if when not having the database of described tone color, in the operation of described voice adaptation, the stage S18 in the voice tamber data storehouse identified described in renewal.
Preferably also comprise: in the described S17 stage, check in the described S11 stage, after contrasting with described respective acoustic model, whether there is the stage S19 of the described acoustic model with mxm.; And in the described S19 stage, when to there is not the described acoustic model with mxm. in all acoustic models if be checked through, record does not have and the stage S20 of the described voice class that inputs in the described S1 stage like the acoustic model of tone color.
Preferably also comprise: in the described apparatus control in described S15 stage, the zone user name in advance set by speech recognition is designated as the stage S21 of control area user name; And judge in the described S8 stage, the described recognition result of certification be the stage S22 of regional choice instruction or apparatus control instruction; And in the described S22 stage, if when being judged to be regional choice instruction, described zone user name is changed the stage S23 being designated as described control area user name; And in the described S22 stage, if when being judged to be apparatus control instruction, control the stage S24 of described control area user name affiliated area machine.
Further, be preferably also included in the described S23 stage, after the user name of change memory described control area, after a certain time, described set zone user name be restored to the stage S25 of described control area user name.
Further, preferably also comprise: in the described S9 stage, during described recognition result in the S8 stage described in not certification, divide calculation level separately according to the speech category that the described S10 stage classifies, increase the stage S26 of various types of described calculation level; And in the described S26 stage, if when each calculation level first arrives fixed arbitrary value, if then change to noise pattern, if then change to the stage S27 of refusal pattern during described forte during the described off beat of the state of cognitron; And in the described S4 stage, if when described silent sound calculation level first arrives determined threshold value, become the stage S28 of basic model.
Preferably also comprise: according to the pattern that the described S28 stage changes, if during refusal pattern, the identification of all instructions of refusal except described basic model transformation directive, if during noise pattern, reduce artificially input the power of voice, control the noise belonging to the second grade in the described S10 stage, if during basic model, carry out the normal stage S29 identified; And under the described refusal mode state in described S8 stage, if when the identification of first fixed basic model release command is certified, be the stage S30 of described basic model by the mode altering of described speech recognition machine.
The invention has the beneficial effects as follows: the pattern of speech recognition machine can change according to surrounding condition, instant processed voice identification, reduce the wrong identification to non-instruction voice, the discrimination of instruction is improved in noisy environment, automatically the tone color of the voice person do not had in speech recognition database can be adapted to, record acoustic model also identifies it, thus raising discrimination, the instruction of the same race in multiple region can be unified, reduce various instruction, the accuracy rate of speech recognition is improved greatly, effectively reduces wrong identification simultaneously.
Accompanying drawing explanation
Fig. 1 is the general introduction structural drawing of the domestic network system of embodiment of the present invention;
Fig. 2 is the overall flow figure of the audio recognition method of embodiment of the present invention;
Fig. 3 is the detail flowchart of the voice recognition processing process of Fig. 2;
Fig. 4 be Fig. 2 first time recognition result processing procedure detail flowchart;
Fig. 5 is the detail flowchart that process is carried out in the second time recognition result process of Fig. 2;
Fig. 6 is the detail flowchart of the patten transformation processing procedure of Fig. 2;
Fig. 7 is the process flow diagram of the apparatus control process according to the change of control user name of embodiment of the present invention.
Embodiment
Be described principle of the present invention and feature below in conjunction with accompanying drawing, example, only for explaining the present invention, is not intended to limit scope of the present invention.
Fig. 1 is the general introduction structural drawing of the domestic network system of embodiment of the present invention.As shown in Figure 1, speech recognition domestic network system of the present invention can arrange speech recognition machine in each region, and each speech recognition machine is connected by the wired or wireless controller with controlling devices such as throwing light on.Controller carrys out control machine according to the operation of user, analyzes the instruction transmitted by speech recognition machine, thus carrys out control machine.The speech recognition machine in each region has intrinsic position user name, and controller analyzes above-mentioned user name, and controls the machine on corresponding speech recognition machine position set in affiliated area.
Fig. 2 is the overall flow figure of the audio recognition method of embodiment of the present invention.As shown in Figure 2, after system comes into operation, speech recognition machine can receive inputted voice 200 immediately constantly.The power of the voice 201 of lasting input and the duration of a sound and set arbitrary value compare by speech recognition machine, and carry out classification 202 according to silent sound, off beat, forte three kinds.Speech recognition machine, according to classified speech category, does not process for silent sound, carries out voice recognition processing for off beat and forte.
When speech recognition machine carries out identifying processing, first time can be divided into identify 203 and second time identification 205.After carrying out first time identification 203, process 204 is carried out to its recognition result, if when first time identifies that the recognition result of 203 conforms to set condition, then carry out second time and identify 205; If do not conform to, then stop identifying, and be converted to phonetic entry standby phase 200.Recognition result satisfies condition for the first time, carry out second time and identify 205, and be converted to basic model 206 according to second time recognition result control machine or by speech recognition mode, in the stage 202 that voice are classified, according to classified voice, be noise pattern or refusal pattern 207 by the State Transferring of speech recognition machine.
Fig. 3 is the detail flowchart of the voice recognition processing process of Fig. 2.Fig. 3 illustrate in details the voice for inputting in phonetic entry standby 200, by type classification, and judges whether the processing procedure of carrying out voice recognition processing according to kind.Speech recognition machine in speech recognition domestic network system in Fig. 2 can receive inputted voice 200 immediately, and carries out classification 201 according to the power of voice and the duration of a sound.
As shown in Figure 3, be described in detail with regard to speech category classification process.First the power 301 of voice and the duration of a sound 302 of voice is measured.Measured speech volume and the duration of a sound and set arbitrary value compare by speech recognition machine, and carry out classification 303 by silent sound, off beat, forte three kinds to voice.Illustrate, if when the speech volume measured is less than 50dB, this speech category is divided to silent sound class; If the speech volume measured is 50 to 60dB, when the duration of a sound of voice is more than 2 seconds, these voice divide to off beat class; If speech volume is at more than 60dB, and the duration of a sound more than 2 seconds time, these voice divide to forte class.
The kind of voice is divided into silent sound, off beat, forte.According to the kind of voice, if silent sound is then without the need to carrying out voice recognition processing, while the silent sound calculation level 306 of increase, continue to maintain phonetic entry holding state 200.If when the silent sound calculation level increased arrives set arbitrary value 307, speech recognition machine confirms current speech recognition mode, if during basic model, then after silent sound calculation level initialization 310, maintain phonetic entry holding state; If when noise pattern 308 or refusal mode 3 09, speech recognition mode is changed to basic model 311, after silent sound calculation level initialization 310, maintain phonetic entry holding state 200.If when the kind of voice is judged as off beat class or forte class, after silent sound calculation level initialization 310, carry out first time identification 304.
Fig. 4 is the detail flowchart of first time recognition result process 204 process of Fig. 2.As shown in Figure 4, after carrying out first time identification 203, the non-refusal word of the result identified or refusal statement, and when belonging to set instruction, with the logarithm of each acoustic model remembered in first time identifying processing like number than contrasting, tell the acoustic model 402 with mxm..Remember and identify that the logarithm of the acoustic model used in 203 is specific as follows like the process of number ratio in first time: first before speech recognition domestic network system starts, in memory size limits, copy multiple acoustic model according to the number of voice person, and prepare multiple acoustic model; Next, after starting to operate speech recognition domestic network system, in the process of carrying out first time identification 203, after identification particular words, the logarithm that memory is transmitted to the affiliated pattern of virtual machine compares like number.
For example, in speech recognition process, after each word of identification, after dummy node (dummynode), further expansion is other words.As: " baobao " → dummy → " master bedroom " → dummy → " turning on light ".On each dummy node, calculate the probable value of all words towards respective direction respectively, and the word with maximum probability value is designated as identification candidate target.In the process, before speech recognition domestic network system starts operation, if copy 3 acoustic models and cut-and-dried words, dummy node then also exists the ending phoneme from " baobao ", " baobao2 ", " baobao3 ", as " ao-b+ao ", " ao2-b2-ao2 ", " ao3-b3-ao3 " is towards the node of each dummy node, and calculate about the logarithm of above-mentioned each node is like number ratio, and remember that now calculated logarithm is like number ratio.
First time identifies that the result of 203 belongs to set instruction, as the process illustrated by above-mentioned citing, by the logarithm remembered like number than in, when being determined 402 containing the acoustic model with mxm., there is the highest logarithm in second time identification database and be logged internal memory 403 like the database of number ratio.In the process of 403, second time is carried out to the database logined and identify 205.If first time is when identifying the unvested instruction of result of 203, speech recognition machine is then abandoned second time and is identified, and removes internal memory, is transformed into phonetic entry holding state 200 simultaneously, carries out relevant operation.
Fig. 5 is the detail flowchart that process is carried out in the second time recognition result process 206 of Fig. 2.As shown in Figure 5, identifying in the process of carrying out for the second time, as described in Figure 4, the logarithm of each acoustic model calculated in first time identifying like number than in, the acoustic model with mxm. is logged, and for database that second time identifies.As mentioned above, after terminating second time identification, when processing and identification result, first check whether recognition result belongs to set instruction, and check whether it belongs to refusal word or statement 501.
If recognition result belongs to refusal word or statement, after removing internal memory, get back to phonetic entry holding state 200.If recognition result belongs to instruction, then judge whether to there is the database 502 similar with voice person tone color.If the acoustic model of all speech databases is all the same, then judge without the speech data similar with voice person tone color, when there is the acoustic model of mxm. if exist, then it is determined that the presence of the speech data similar with voice person tone color.If when recognition result belongs to instruction, store voice 503, and the voice stored by using, simultaneously because identifying in 205 the adaptation 504 carried out acoustic model in second time, thus upgrade the acoustic model 505 approximate with the tone color of voice person further.
Fig. 6 is the detail flowchart of patten transformation process 207 process of Fig. 2.As shown in Figure 6, in second time recognition result process 206 stage, certification recognition result 601 is judged whether.If not certification recognition result, then differentiating that the stage 202 of speech category differentiates that the voice of classifying belong to off beat or forte.
If voice belong to off beat, then whether the pattern that differentiation speech recognition machine is current is noise pattern 610, if noise pattern, then removes the internal memory of use, enters phonetic entry standby process 200 stage.If the voice inputted are off beat, but during the non-noise pattern of the pattern of current speech cognitron, then increase noise calculation level 611, and differentiate whether the calculation level increased arrives set arbitrary value 612, if when arriving arbitrary value, be then noise state 613 by the Status Change of speech recognition machine.
If speech recognition machine changes to noise pattern, speech recognition machine regulates the volume inputted, and the voice of certain volume are considered as silent sound, and processes the voice exceeding above-mentioned certain volume.In addition, not certified at second time recognition result, and when the kind of voice is forte, then confirm whether the pattern of current speech cognitron is refusal pattern 614.
If during refusal pattern, then remove the internal memory used in speech recognition machine, and carry out the standby process 200 of phonetic entry.If not during refusal pattern, then increase refusal calculation level 615.Checking whether the refusal calculation level that increases arrives set arbitrary value 616, namely determine arbitrary value if arrive, is then refusal pattern 617 by the patten transformation of speech recognition machine.
When speech recognition machine changes to refusal pattern, first time recognition result processing stage 204 and second time recognition result processing stage in apparatus control instruction be not then identified certification, enter the refusal stage simultaneously.When recognition result processing stage of second time recognition result in 206 is certified, as shown in Figure 6, when recognition result is certified, confirm whether the pattern of current speech cognitron is refusal pattern 602.
If refusal pattern, check whether currently identified instruction is the instruction 603 changing basic model.If currently identified instruction be conversion basic model instruction time, then present mode is changed to basic model 604, and enter phonetic entry standby processing stage 200.If second time recognition result is certified, but during the non-refusal pattern of present mode, confirm that the instruction identified is apparatus control instruction, or place transformation directive 605,607.
If during apparatus control instruction, then control correlation machine 606, and get back to phonetic entry standby processing stage 200.If during the transformation directive of place, identify that the control area user name of the speech recognition machine in operation changes note on determined venue users name 608.As above changed control area user name after certain hour, then resets in the user name of setting area.
Fig. 7 is the process flow diagram of the apparatus control process according to the change of control user name of embodiment of the present invention.As shown in Figure 7, when speech recognition home network starts to start, before carrying out speech recognition process, set zone user name is designated as control area user name 701, such as, the control area user name that speech recognition machine is remembered is as shown in table 1 to be set.
[table 1]
| Region | User name |
| Parlor | 00 |
| Room 1 | 01 |
| Room 2 | 02 |
| Room 3 | 03 |
According to remembered control area user name, voice identification result when operating machines 704, confirms the user name 705 remembered, and selects and control the machine 709,710 that will control.If when voice identification result is regional choice instruction, confirm to change which region 706, control area user name changes to parlor or room 1707 by phonetic order, 708, above-mentioned changed control area user name can not get the confirmation of speech recognition within a certain period of time and continuous standby time 702, when holding state reaches the standby calculation level of specifying in advance 703, the function 701 of getting back to control area user name can be considered.
If during without control area user name, assuming that parlor, master bedroom and little crouching arrange electric light respectively, when user will turn on parlor electric light, need " turning on parlor lamp ", when master bedroom lamp will be turned on, need " turning on master bedroom lamp ", when little sleeping lamp will be turned on, need " turning on little sleeping lamp ", so need different instructions according to different regions.Thus, the kinds of machine that speech recognition machine controls is more, and its relevant instruction then also can increase thereupon.
Like this, speech recognition machine just needs the many recognition instructions of input and word, thus causes the internal memory of speech recognition machine and function also to increase thereupon, when constructing speech recognition home network, needs more expense and research.
But as disclosed in the present invention described in method, for needing memory in addition and the region of control area user name, if this region is the words in the region not arranging speech recognition machine, first identify change region instruction, then the instruction that will control is identified, so all multiple instruction integrations can be got up, both can reduce the expense of great number, also can improve that it is technical and explorative widely.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (9)
1. be applied to an audio recognition method for home network, it is characterized in that, described audio recognition method comprises: the instant stage S1 receiving phonetic entry; Measure the stage S2 of the volume of described voice and the duration of a sound of voice inputted in the above-mentioned S1 stage; By the duration of a sound of the volume of described voice that measures in the described S2 stage and voice compared with arbitrary value, and by the stage S3 that silent sound, off beat, forte are classified to described voice;
In the described S3 stage, according to the power formulated in advance, described voice are divided into Three Estate, if during the minimum the first estate of the volume of the voice measured in the described S2 stage, then silent sound class assigned in described voice; If the volume of described voice belongs to the second grade higher than described the first estate, and when the duration of a sound of described voice is arbitrary value, then off beat class assigned in described voice; If the volume of described voice is the highest tertiary gradient, and when the duration of a sound of described voice meets arbitrary value, then the S10 stage of forte class assigned in described voice;
According to the classification in the described S3 stage, if silent sound, then calculate the duration of silent sound, and continue to maintain phonetic entry holding state, if off beat or forte, then carry out the stage S4 of voice recognition processing; According to the voice recognition processing that the described S4 stage carries out, carry out the stage S5 formulating unit voice recognition processing; According to the result in described S5 stage, determine whether voice recognition processing is carried out to overall voice, the stage S6 that row relax of going forward side by side prepares; According to the described S6 stage, perform the voice recognition processing of overall voice if determine, then overall voice are carried out to the stage S7 of voice recognition processing; According to the result in described S7 stage, determine the result of whether certification voice recognition processing, and control machine, carry out the stage S8 identifying voice adaptation; In the described S6 stage, when speech recognition not being carried out to described overall voice, or according to the described S8 stage, during the recognition result of voice described in not certification, then according to the Classification of Speech determined in the described S3 stage, if during off beat, be then converted to noise pattern, if during forte, be then converted to the stage S9 of refusal pattern.
2. the audio recognition method being applied to home network according to claim 1, it is characterized in that, described audio recognition method also comprises: in the process that the voice recognition processing of the described formulation unit in described S5 stage is carried out, if during identidication key, the log-likelihood ratio described key word most end phoneme being transmitted to virtual machine contrasts with key word and acoustic model corresponding to key word most end phoneme, and memory has the stage S11 of the acoustic model of mxm.; And
In the described S6 stage, institute's speech recognition result with to deposit instruction consistent, and determines to carry out to described overall voice the stage S12 that second time identifies; And
In the described S12 stage, when determining to carry out described second time speech recognition, the described acoustic model remembered is changed to the preparatory stage S13 of second time recognition mode in the described S11 stage; And
In the described S13 stage, the acoustic model with mxm. is used for database, and carries out the stage S14 of second time speech recognition.
3. the audio recognition method being applied to home network according to claim 2, it is characterized in that, described audio recognition method also comprises: in the described S8 stage, if described second time voice identification result is certified, according to the stage S15 of described authentication result control machine; And
In the described S8 stage, store described overall voice, and carry out the adaptation operation of stored voice, upgrade the stage S16 with the acoustic model of the tamber characteristic of the voice person of pronunciation.
4. the audio recognition method being applied to home network according to claim 3, it is characterized in that, described audio recognition method also comprises: in the described S16 stage, checks whether the stage S17 of the database of the tamber characteristic of the voice person with described pronunciation; And
In the described S16 stage, if when not having the database of the tamber characteristic of the voice person of described pronunciation, in the operation of described voice adaptation, upgrade the stage S18 of the database of the tamber characteristic of the voice person of the pronunciation identified.
5. the audio recognition method being applied to home network according to claim 4, it is characterized in that, described audio recognition method also comprises: in the described S17 stage, check in the described S11 stage, with described with key word and after acoustic model corresponding to key word most end phoneme contrast, whether there is the stage S19 of the described acoustic model with mxm.; And
In the described S19 stage, when to there is not the described acoustic model with mxm. in all acoustic models if be checked through, record does not have and the stage S20 of the described voice class that inputs in the described S1 stage like the acoustic model of tone color.
6. the audio recognition method being applied to home network according to claim 3, it is characterized in that, described audio recognition method also comprises: in the described control machine in described S15 stage, the zone user name in advance set by speech recognition machine is designated as the stage S21 of control area user name; And
Judge in the described S8 stage, the described recognition result of certification be the stage S22 of regional choice instruction or apparatus control instruction; And in the described S22 stage, if when being judged to be regional choice instruction, described zone user name is changed the stage S23 being designated as described control area user name; And
In the described S22 stage, if when being judged to be apparatus control instruction, control the stage S24 of described control area user name affiliated area machine.
7. the audio recognition method being applied to home network according to claim 6, it is characterized in that, described audio recognition method also comprises: in the described S23 stage, after the user name of change memory described control area, after a certain time, described changed zone user name is restored to the stage S25 of described control area user name.
8. the audio recognition method being applied to home network according to claim 1, it is characterized in that, described audio recognition method also comprises: in the described S9 stage, during described recognition result in the S8 stage described in not certification, divide calculation level separately according to the speech category that the described S10 stage classifies, increase the stage S26 of various types of described calculation level; And
In the described S26 stage, if when each calculation level first arrives fixed arbitrary value, if then change to noise pattern, if then change to the stage S27 of refusal pattern during described forte during the described off beat of the state of speech category; And in the described S4 stage, if when described silent sound calculation level first arrives determined threshold value, become the stage S28 of basic model.
9. the audio recognition method being applied to home network according to claim 8, it is characterized in that, described audio recognition method also comprises: according to the pattern that the described S28 stage changes, if during refusal pattern, refuse the identification of all instructions except described basic model transformation directive, if during noise pattern, reduce artificially input the power of voice, control the noise belonging to the second grade in the described S10 stage, if during basic model, carry out the normal stage S29 identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110065918.0A CN102682767B (en) | 2011-03-18 | 2011-03-18 | Speech recognition method applied to home network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110065918.0A CN102682767B (en) | 2011-03-18 | 2011-03-18 | Speech recognition method applied to home network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102682767A CN102682767A (en) | 2012-09-19 |
| CN102682767B true CN102682767B (en) | 2015-04-08 |
Family
ID=46814583
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110065918.0A Expired - Fee Related CN102682767B (en) | 2011-03-18 | 2011-03-18 | Speech recognition method applied to home network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102682767B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106981290B (en) * | 2012-11-27 | 2020-06-30 | 威盛电子股份有限公司 | Voice control device and voice control method |
| CN103885529A (en) * | 2012-12-19 | 2014-06-25 | Cs-Ucamp株式公司 | Wall-pad system by wireless access point and panel computer and movement method of wall-pad system |
| CN109379499A (en) * | 2018-11-20 | 2019-02-22 | 北京千丁互联科技有限公司 | A kind of voice call method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1391210A (en) * | 2001-05-11 | 2003-01-15 | 皇家菲利浦电子有限公司 | Method for training or adapting to phonetic recognizer |
| CN1545693A (en) * | 2001-08-22 | 2004-11-10 | �Ҵ���˾ | Intonation generating method, speech synthesizing device by the method, and voice server |
| CN101046958A (en) * | 2006-03-29 | 2007-10-03 | 株式会社东芝 | Apparatus and method for speech processing |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10313310A1 (en) * | 2003-03-25 | 2004-10-21 | Siemens Ag | Procedure for speaker-dependent speech recognition and speech recognition system therefor |
-
2011
- 2011-03-18 CN CN201110065918.0A patent/CN102682767B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1391210A (en) * | 2001-05-11 | 2003-01-15 | 皇家菲利浦电子有限公司 | Method for training or adapting to phonetic recognizer |
| CN1545693A (en) * | 2001-08-22 | 2004-11-10 | �Ҵ���˾ | Intonation generating method, speech synthesizing device by the method, and voice server |
| CN101046958A (en) * | 2006-03-29 | 2007-10-03 | 株式会社东芝 | Apparatus and method for speech processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102682767A (en) | 2012-09-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7769588B2 (en) | Spoken man-machine interface with speaker identification | |
| EP3353677B1 (en) | Device selection for providing a response | |
| CN105009204B (en) | Speech recognition power management | |
| CN108231079A (en) | For the method, apparatus, equipment and computer readable storage medium of control electronics | |
| CN104580699B (en) | Acoustic control intelligent terminal method and device when a kind of standby | |
| CN107210040A (en) | The operating method of phonetic function and the electronic equipment for supporting this method | |
| CN107767861A (en) | voice awakening method, system and intelligent terminal | |
| TW200826064A (en) | Speech recognition method and system with intelligent classification and adjustment | |
| CN102999161A (en) | Implementation method and application of voice awakening module | |
| CN107729433B (en) | Audio processing method and device | |
| CN111862951B (en) | Voice endpoint detection method and device, storage medium and electronic equipment | |
| CN102682767B (en) | Speech recognition method applied to home network | |
| JP2000310999A (en) | Facilities control system | |
| CN111178081A (en) | Semantic recognition method, server, electronic device and computer storage medium | |
| CN107742520A (en) | Sound control method, apparatus and system | |
| US11475876B2 (en) | Semantic recognition method and semantic recognition device | |
| US11240641B1 (en) | Automated device grouping | |
| CN108870650A (en) | A kind of air-conditioning and a kind of control method of air-conditioning | |
| CN113516975A (en) | Intelligent household voice-operated switch system and control method | |
| CN111477226A (en) | Control method, intelligent device and storage medium | |
| Fezari et al. | New speech processor and ultrasonic sensors based embedded system to improve the control of a motorised wheelchair | |
| CN116386623A (en) | Voice interaction method of intelligent equipment, storage medium and electronic device | |
| US11244677B2 (en) | Voice control system and method for moving a column | |
| EP1387350A1 (en) | Spoken man-machine interface with speaker identification | |
| CN210575088U (en) | Voice recognition household appliance control device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150408 Termination date: 20170318 |