Summary of the invention
In order to solve the above technical problems, an embodiment of the present invention is intended to provide a kind of voice information identification method and equipment, solutionDetermined voice messaging identifying schemes in the prior art calculation amount is larger and problem that difficulty complicated for operation is larger, reduce languageThe difficulty of message breath identification, reduces calculation amount；Meanwhile improving the experience effect of user.
The technical scheme of the present invention is realized as follows:
A kind of voice information identification method, which comprises
Obtain voice stream information to be identified；
The voice stream information to be identified is analyzed, extract the voice stream information corresponding first to be identified toAmount；
The primary vector is subjected to segmentation arrangement, obtains secondary vector；
The primary vector is classified according to default principle of classification to obtain third vector；
According to the relationship between third vector described in the secondary vector of each segmentation and each classification, by described wait knowOther voice stream information is matched with user.
Optionally, described that the primary vector is subjected to segmentation arrangement, obtain secondary vector, comprising:
According to the play time of the voice stream information to be identified, according to the prefixed time interval by the primary vectorSegmentation arrangement is carried out, the secondary vector is obtained.
It is optionally, described that the primary vector is classified according to default principle of classification to obtain third vector, comprising:
Principal component analysis is carried out to the primary vector, obtains the 4th vector；
Taxonomic revision is carried out to the primary vector according to the 4th vector, obtains the third vector.
Optionally, it is described according to the 4th vector to the primary vector carry out taxonomic revision, obtain the third toAmount, comprising:
Variation Bayes gauss hybrid models cluster is carried out to the primary vector according to the 4th vector, is obtained describedThird vector.
Optionally, the pass between the third vector according to the secondary vector of each segmentation and each classificationSystem, the voice stream information to be identified is matched with subscriber identity information, comprising:
Each secondary vector in each segmentation is matched with each third vector in each classification；
If each third vector in each secondary vector and each classification in each segmentation exactly matches,Then voice stream information to be identified corresponding to the secondary vector in each segmentation carries out voiceprint identification respectively, obtains instituteState the identity information of the corresponding user of voice stream information to be identified.
Optionally, the method also includes:
If each incomplete of third vector in each secondary vector and each classification in each segmentationMatch, then in the secondary vector segmentation carry out again segment processing simultaneously use Wei Tebi algorithm to the third vector intoRow reclassifies, the secondary vector in each section until being segmented again with reclassify after each described theThe exact matching of three vectors；
Voice stream information to be identified corresponding to the secondary vector in each segmentation carries out voiceprint identification respectively,Obtain the identity information of the corresponding user of the voice stream information to be identified.
A kind of voice messaging identification equipment, the equipment include: first acquisition unit, second acquisition unit, third acquisitionUnit and processing unit, in which:
The first acquisition unit, for obtaining voice stream information to be identified；
The first acquisition unit is also used to analyze the voice stream information to be identified, extracts described to be identifiedThe corresponding primary vector of voice stream information；
The second acquisition unit obtains secondary vector for the primary vector to be carried out segmentation arrangement；
The third acquiring unit, for according to default principle of classification to the primary vector classified to obtain third toAmount；
The processing unit, between the third vector according to the secondary vector of each segmentation and each classificationRelationship, the voice stream information to be identified is matched with user.
Optionally, the second acquisition unit is specifically used for:
According to the play time of the voice stream information to be identified, according to the prefixed time interval by the primary vectorSegmentation arrangement is carried out, the secondary vector is obtained.
Optionally, the third acquiring unit includes: analysis module and first processing module, in which:
The analysis module obtains the 4th vector for carrying out principal component analysis to the primary vector；
The first processing module is obtained for carrying out taxonomic revision to the primary vector according to the 4th vectorThe third vector.
Optionally, the first processing module is specifically used for:
Variation Bayes gauss hybrid models cluster is carried out to the primary vector according to the 4th vector, is obtained describedThird vector.
Optionally, the processing unit includes: matching module and Second processing module, in which:
The matching module, for by each secondary vector in each segmentation and each described in each classificationThird vector is matched；
The Second processing module, if in each segmentation each secondary vector with it is each in each classificationThird vector exact matching, then respectively to the corresponding voice stream information to be identified of the secondary vector in each segmentation intoThe identification of row voiceprint, obtains the identity information of the corresponding user of the voice stream information to be identified.
Optionally, the processing unit further include: third processing module and fourth processing module, in which:
The third processing module, if in each segmentation each secondary vector with it is each in each classificationThe third vector Incomplete matching then carries out again segment processing to the segmentation in the secondary vector and uses Wei Tebi simultaneouslyAlgorithm reclassifies the third vector, the secondary vector in each section until being segmented again with againEach third vector exact matching after classification；
The fourth processing module, for respectively to the corresponding voice flow to be identified of the secondary vector in each segmentationInformation carries out voiceprint identification, obtains the identity information of the corresponding user of the voice stream information to be identified.
Voice information identification method and equipment provided by the embodiment of the present invention, available voice flow letter to be identifiedBreath, and voice stream information to be identified is analyzed, the corresponding primary vector of voice stream information to be identified is extracted, then by firstVector carries out segmentation arrangement, obtains secondary vector, while being classified to obtain third to primary vector according to default principle of classificationVector, finally, being believed voice flow to be identified according to the relationship between the secondary vector of each segmentation and each classification third vectorBreath is matched with user, in this way, can be obtained according to the corresponding vector of voice stream information to be identified and voice flow to be identifiedThe user of information matches, solve voice messaging identifying schemes in the prior art calculation amount is larger and difficulty complicated for operation compared withBig problem reduces the difficulty of voice messaging identification, reduces calculation amount；Meanwhile improving the experience effect of user.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, completeSite preparation description.
The embodiment of the present invention provides a kind of voice information identification method, and shown referring to Fig.1, this method includes following stepIt is rapid:
Step 101 obtains voice stream information to be identified.
It can be equipment is identified by voice messaging to realize specifically, step 101 obtains voice stream information to be identified.ToIdentification voice stream information can be the voice messaging for needing to carry out speech recognition that user is input to electronic equipment, need to identifyVoice messaging, which can be, to be acquired by the voice collector such as microphone etc. of electronic equipment.
Step 102 analyzes voice stream information to be identified, extract voice stream information to be identified corresponding first toAmount.
Specifically, step 102 analyzes voice stream information to be identified, voice stream information to be identified corresponding is extractedOne vector, which can be, to be identified equipment by voice messaging to realize；The voice stream information identified can will be needed in initial fragment itemVector is extracted under part and is standardized to obtain primary vector.
Primary vector is carried out segmentation arrangement by step 103, obtains secondary vector.
Specifically, primary vector is carried out segmentation arrangement by step 103, obtain secondary vector and can be to be identified by voice messagingEquipment is realized.Wherein, the segmentation of primary vector can be uniformly is segmented according to the period pre-set,It can be and segmentation heterogeneous is carried out according to specific demand.
Step 104 classifies primary vector according to default principle of classification to obtain third vector.
Specifically, step 104 according to default principle of classification to primary vector classified to obtain third vector can be byVoice messaging identifies equipment to realize.
Step 105, according to the relationship between the secondary vector and each classification third vector of each segmentation, by language to be identifiedSound stream information is matched with user.
It, will be to specifically, step 105 is according to the relationship between the secondary vector and each classification third vector of each segmentationIdentification voice stream information and user match can be by voice messaging identification equipment and realize；After can comparing segmentationThe matching relationship between third vector after the obtained secondary vector in each section and classification in every one kind, and according to comparingAs a result the information for needing the corresponding user of voice messaging identified is obtained, realizes voice stream information to be identified and the matching of user.
Voice information identification method provided by the embodiment of the present invention, available voice stream information to be identified, and it is rightVoice stream information to be identified is analyzed, and the corresponding primary vector of voice stream information to be identified is extracted, then by primary vector intoRow segmentation arranges, and obtains secondary vector, while being classified to obtain third vector to primary vector according to default principle of classification, mostAfterwards, it according to the relationship between the secondary vector of each segmentation and each classification third vector, by voice stream information to be identified and usesFamily is matched, in this way, can be obtained according to the corresponding vector of voice stream information to be identified and voice stream information to be identifiedThe user matched, the calculation amount for solving voice messaging identifying schemes in the prior art are larger and difficulty complicated for operation is biggish asksTopic reduces the difficulty of voice messaging identification, reduces calculation amount；Meanwhile improving the experience effect of user.
The embodiment of the present invention provides a kind of voice information identification method, and referring to shown in Fig. 2, this method includes following stepIt is rapid:
Step 201, voice messaging identification equipment obtain voice stream information to be identified.
Step 202, voice messaging identification equipment analyze voice stream information to be identified, extract voice flow letter to be identifiedCease corresponding primary vector.
Specifically, primary vector, which can be, carries out i vector to the voice stream information that needs identify under the conditions of initial fragmentExtraction and standardization obtain；Such as the time that the voice messaging that can be identified as required plays, and as unit of every 1 secondIt carries out initial fragment and obtains the corresponding primary vector of voice messaging for needing to identify.
Step 203, voice messaging identify equipment according to the play time of voice stream information to be identified, according between preset timeEvery primary vector is carried out segmentation arrangement, secondary vector is obtained.
Wherein, which can be user's broadcasting according to the voice messaging for needing to identify in concrete application sceneIt puts duration, need the factors preset one such as the quantity of corresponding user in the voice messaging that identifies, the successful ratio of identificationA time interval, such as can be the voice letter identified as required using one minute or five minutes constant durations as unitActual play time and sequence are ceased, the voice messaging in each minute or every five minutes durations is one section, every after segmentationIt is secondary vector that one section, which needs the vector set of its corresponding vector composition of the voice messaging identified,.
It should be noted that said in the present embodiment be primary vector be segmented according to prefixed time interval it is wholeReason obtain secondary vector, equally can according to different time intervals to primary vector carry out segmentation arrange obtain second toAmount, specific segmentation scheme can be determined according to actual application scenarios.
Step 204, voice messaging identification equipment carry out principal component analysis to primary vector, obtain the 4th vector.
Specifically, carrying out principal component analysis to primary vector can be based on certain factor analysis primary vector, specificallyThe realization process of principal component analysis be referred to the related art scheme about principal component analysis in the prior art.
Step 205, voice messaging identification equipment according to the 4th vector to primary vector carry out taxonomic revision, obtain third toAmount.
Specifically, can be on the basis of the 4th vector to what primary vector was classified, primary vector is mapped to oftenIn one the 4th vector, classified to obtain third vector later according to actual mapping result.
Step 206, according to the relationship between the secondary vector and each classification third vector of each segmentation, by language to be identifiedSound stream information is matched with user.
It should be noted that the explanation in the present embodiment with same steps in other embodiments or concept, is referred to itDescription in its embodiment, details are not described herein again.
Voice information identification method provided by the embodiment of the present invention, available voice stream information to be identified, and it is rightVoice stream information to be identified is analyzed, and the corresponding primary vector of voice stream information to be identified is extracted, then by primary vector intoRow segmentation arranges, and obtains secondary vector, while being classified to obtain third vector to primary vector according to default principle of classification, mostAfterwards, it according to the relationship between the secondary vector of each segmentation and each classification third vector, by voice stream information to be identified and usesFamily is matched, in this way, can be obtained according to the corresponding vector of voice stream information to be identified and voice stream information to be identifiedThe user matched, the calculation amount for solving voice messaging identifying schemes in the prior art are larger and difficulty complicated for operation is biggish asksTopic reduces the difficulty of voice messaging identification, reduces calculation amount；Meanwhile improving the experience effect of user.
The embodiment of the present invention provides a kind of voice information identification method, and referring to shown in Fig. 3, this method includes following stepIt is rapid:
Step 301, voice messaging identification equipment obtain voice stream information to be identified.
Step 302, voice messaging identification equipment analyze voice stream information to be identified, extract voice flow letter to be identifiedCease corresponding primary vector.
Step 303, voice messaging identify equipment according to the play time of voice stream information to be identified, according between preset timeEvery primary vector is carried out segmentation arrangement, secondary vector is obtained.
Step 304, voice messaging identification equipment carry out principal component analysis to primary vector, obtain the 4th vector.
Step 305, voice messaging identification equipment carry out variation Bayes Gaussian Mixture to primary vector according to the 4th vectorModel tying obtains third vector.
Specifically, voice messaging identification equipment can be one coordinate system of formation on the basis of the 4th vector, by eachPrimary vector is mapped in the coordinate system of the 4th vector formation, is clustered later using variation Bayes gauss hybrid models to mappingPrimary vector afterwards is classified, and third vector is obtained.
Step 306, voice messaging identification equipment by each segmentation each secondary vector with it is each in each classificationThird vector is matched.
Specifically, each that all vectors being segmented in each obtained secondary vector and classification can be obtainedInstitute's directed quantity in third vector carries out the judgement of matching degree, wherein the corresponding third vector of a secondary vector.
It should be noted that step 306 by each segmentation each secondary vector and each classification in each thirdVector can select to execute step 307 or step 308~309 according to matching result after being matched, if each segmentationIn each secondary vector and each classification in each third vector exact matching then follow the steps 307, if each segmentationIn each secondary vector and each classification in each third vector Incomplete matching then follow the steps 308~309；
If each third vector in each secondary vector and each classification in step 307, each segmentation exactly matches,Then voice stream information to be identified corresponding to the secondary vector in each segmentation carries out vocal print letter to voice messaging identification equipment respectivelyBreath identification, obtains the identity information of the corresponding user of voice stream information to be identified.
Specifically, if institute's directed quantity in secondary vector in each segmentation with it is all in corresponding third vectorVector all matches, and illustrates that each secondary vector in each segmentation and each third vector in each classification exactly match, thisWhen can consider that the segmentation of primary vector be more accurately, it is believed that one segmentation in the corresponding needs of secondary vectorIt is the voice messaging of a user that the voice messaging of identification is corresponding, can directly be believed according to the vocal print feature of each user itselfBreath etc. matches the corresponding voice stream information to be identified of secondary vector with the information of user, obtains the body of its corresponding userPart information.Wherein, vocal print feature information may include that tone color, tone, sound quality, volume of user etc. being capable of unique identification usersCharacteristic information.
If each third vector incomplete in each secondary vector and each classification in step 308, each segmentationMatch, then voice messaging identification equipment carries out again segment processing to the segmentation in secondary vector and uses Wei Tebi algorithm to the simultaneouslyThree vectors are reclassified, the secondary vector in each section until being segmented again with reclassify after every 1 theThe exact matching of three vectors.
Specifically, if the vector in secondary vector in each segmentation exist with the vector in corresponding third vector notMatched vector illustrates that each secondary vector in each segmentation and each third vector in each classification are not completeMatch, need to be segmented segmentation result again at this time, while needing to carry out again classification results using viterbi algorithmWhether classification, the secondary vector after being then segmented more again exactly match with the third vector after reclassifying, ifThere are still unmatched vectors, then continue to be segmented the secondary vector after being segmented again, while to reclassifying itThird vector afterwards using viterbi algorithm reclassify until again be segmented after each section in secondary vector withEach third vector exact matching after reclassifying；Again be segmented after each section in secondary vector with reclassifyAfter each third vector exact matching later, it is believed that the segmentation result being segmented again to primary vector is than calibratedIt is true, it is believed that secondary vector in a segmentation is corresponding need the voice messaging identified it is corresponding be a user languageMessage breath, can according to vocal print feature information of each user itself etc. by the corresponding voice stream information to be identified of secondary vector withThe information of user matches, and obtains the identity information of its corresponding user.
Step 309, voice messaging identify equipment voice flow to be identified corresponding to the secondary vector in each segmentation respectivelyInformation carries out voiceprint identification, obtains the identity information of the corresponding user of voice stream information to be identified.
It should be noted that the explanation in the present embodiment with same steps in other embodiments or concept, is referred to itDescription in its embodiment, details are not described herein again.
The voice information identification method that the embodiment of the present invention provides, available voice stream information to be identified, and treatIdentification voice stream information is analyzed, and is extracted the corresponding primary vector of voice stream information to be identified, is then carried out primary vectorSegmentation arranges, and obtains secondary vector, while being classified to obtain third vector to primary vector according to default principle of classification, mostAfterwards, it according to the relationship between the secondary vector of each segmentation and each classification third vector, by voice stream information to be identified and usesFamily is matched, in this way, can be obtained according to the corresponding vector of voice stream information to be identified and voice stream information to be identifiedThe user matched, the calculation amount for solving voice messaging identifying schemes in the prior art are larger and difficulty complicated for operation is biggish asksTopic reduces the difficulty of voice messaging identification, reduces calculation amount；Meanwhile improving the experience effect of user.
The embodiment of the present invention provides a kind of voice messaging identification equipment 4, and voice messaging identification equipment can be applied toIn a kind of voice information identification method that the corresponding embodiment in Fig. 1~3 provides, referring to shown in Fig. 4, which may include: theOne acquiring unit 41, second acquisition unit 42, third acquiring unit 43 and processing unit 44, in which:
First acquisition unit 41, for obtaining voice stream information to be identified.
First acquisition unit 41 is also used to analyze voice stream information to be identified, extracts voice stream information to be identifiedCorresponding primary vector.
Second acquisition unit 42 obtains secondary vector for primary vector to be carried out segmentation arrangement.
Third acquiring unit 43 obtains third vector for being classified according to default principle of classification to primary vector.
Processing unit 44 will for the relationship between the secondary vector and each classification third vector according to each segmentationVoice stream information to be identified is matched with user.
The voice messaging that the embodiment of the present invention provides identifies equipment, available voice stream information to be identified, and treatsIdentification voice stream information is analyzed, and is extracted the corresponding primary vector of voice stream information to be identified, is then carried out primary vectorSegmentation arranges, and obtains secondary vector, while being classified to obtain third vector to primary vector according to default principle of classification, mostAfterwards, it according to the relationship between the secondary vector of each segmentation and each classification third vector, by voice stream information to be identified and usesFamily is matched, in this way, can be obtained according to the corresponding vector of voice stream information to be identified and voice stream information to be identifiedThe user matched, the calculation amount for solving voice messaging identifying schemes in the prior art are larger and difficulty complicated for operation is biggish asksTopic reduces the difficulty of voice messaging identification, reduces calculation amount；Meanwhile improving the experience effect of user.
Specifically, second acquisition unit 42 is specifically used for executing following steps:
According to the play time of voice stream information to be identified, primary vector be segmented according to prefixed time interval wholeReason, obtains secondary vector.
Further, referring to Figure 5, third acquiring unit 43 includes: analysis module 431 and first processing module 432,In:
Analysis module 431 obtains the 4th vector for carrying out principal component analysis to primary vector.
First processing module 432 obtains third vector for carrying out taxonomic revision to primary vector according to the 4th vector.
Specifically, first processing module 432 is also used to execute following steps:
Variation Bayes gauss hybrid models cluster is carried out to primary vector according to the 4th vector, obtains third vector.
Further, referring to shown in Fig. 6, processing unit 44 includes: matching module 441 and Second processing module 442, in which:
Matching module 441, for by each segmentation each secondary vector and each classification in each third vectorIt is matched.
Second processing module 442, if for each secondary vector in each segmentation and each third in each classificationVector exact matching, then voice stream information to be identified corresponding to the secondary vector in each segmentation carries out voiceprint knowledge respectivelyNot, the identity information of the corresponding user of voice stream information to be identified is obtained.
Further, referring to shown in Fig. 7, processing unit 44 further include: third processing module 443 and fourth processing module 444,Wherein:
Third processing module 443, if for each secondary vector in each segmentation and each third in each classificationVector Incomplete matching, then in secondary vector segmentation carry out again segment processing simultaneously use Wei Tebi algorithm to third toAmount is reclassified, the secondary vector in each section until being segmented again with reclassify after each third toAmount exact matching.
Fourth processing module 444, for voice stream information to be identified corresponding to the secondary vector in each segmentation respectivelyVoiceprint identification is carried out, the identity information of the corresponding user of voice stream information to be identified is obtained.
It should be noted that the interactive process in the embodiment of the present invention between each unit and module, be referred to Fig. 1~The interactive process in a kind of voice information identification method that 3 corresponding embodiments provide, details are not described herein again.
The voice messaging that the embodiment of the present invention provides identifies equipment, available voice stream information to be identified, and treatsIdentification voice stream information is analyzed, and is extracted the corresponding primary vector of voice stream information to be identified, is then carried out primary vectorSegmentation arranges, and obtains secondary vector, while being classified to obtain third vector to primary vector according to default principle of classification, mostAfterwards, it according to the relationship between the secondary vector of each segmentation and each classification third vector, by voice stream information to be identified and usesFamily is matched, in this way, can be obtained according to the corresponding vector of voice stream information to be identified and voice stream information to be identifiedThe user matched, the calculation amount for solving voice messaging identifying schemes in the prior art are larger and difficulty complicated for operation is biggish asksTopic reduces the difficulty of voice messaging identification, reduces calculation amount；Meanwhile improving the experience effect of user.
In practical applications, the first acquisition unit 41, second acquisition unit 42, third acquiring unit 43, processing are singleFirst 44, analysis module 431, first processing module 432, matching module 441, Second processing module 442, third processing module 443It can be by central processing unit (the Central Processing in wireless data sending device with fourth processing module 444Unit, CPU), microprocessor (Micro Processor Unit, MPU), digital signal processor (Digital SignalProcessor, DSP) or field programmable gate array (Field Programmable Gate Array, FPGA) etc. realize.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer programProduct.Therefore, the shape of hardware embodiment, software implementation or embodiment combining software and hardware aspects can be used in the present inventionFormula.Moreover, the present invention, which can be used, can use storage in the computer that one or more wherein includes computer usable program codeThe form for the computer program product implemented on medium (including but not limited to magnetic disk storage and optical memory etc.).
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program productFigure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructionsThe combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programsInstruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produceA raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for realThe device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spyDetermine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram orThe function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that countingSeries of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer orThe instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram oneThe step of function of being specified in a box or multiple boxes.
The foregoing is only a preferred embodiment of the present invention, is not intended to limit the scope of the present invention.