CN103578468B - The method of adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition - Google Patents
The method of adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition Download PDFInfo
- Publication number
- CN103578468B CN103578468B CN201210272154.7A CN201210272154A CN103578468B CN 103578468 B CN103578468 B CN 103578468B CN 201210272154 A CN201210272154 A CN 201210272154A CN 103578468 B CN103578468 B CN 103578468B
- Authority
- CN
- China
- Prior art keywords
- value
- threshold value
- parameter
- confidence threshold
- confidence
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012360 testing method Methods 0.000 claims abstract description 44
- 238000001514 detection method Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 14
- 238000007689 inspection Methods 0.000 claims description 5
- 230000000052 comparative effect Effects 0.000 claims 2
- 238000012549 training Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 8
- 238000013507 mapping Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 230000007423 decrease Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 241000208340 Araliaceae Species 0.000 description 3
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 235000008434 ginseng Nutrition 0.000 description 3
- 238000011897 real-time detection Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011982 device technology Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011895 specific detection Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Electrically Operated Instructional Devices (AREA)
Abstract
The present invention provides the method for adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition.The method is applied in the electronic equipment of a support speech recognition, and the current confidence threshold value of the speech recognition is the first value, and methods described includes：N number of parameter is detected, N number of testing result is obtained, wherein, N is the integer more than or equal to 1；The confidence threshold value is at least adjusted based on a testing result in N number of testing result so that the confidence threshold value is changed into second value from first value, wherein, the second value is the value identical or different with first value.
Description
Technical field
The present invention relates to field of computer technology, more particularly to a kind of confidence coefficient threshold of voice recognition method of adjustment and
Electronic equipment.
Background technology
With the development of electronic device technology, various electronic equipments enter the life of user, as voice is known
The development of other technology, user by Voice command electronic equipment or the scene for carrying out interactive voice with electronic equipment increasingly
Many, the life for giving people brings great convenience.
In the case of Voice command or interactive voice, speech recognition is a critically important step, in the mistake of speech recognition
, it is necessary to confirm to recognition result in journey, will the confidence score of recognition result be compared with confidence threshold value, determine
Whether receive recognition result, when the confidence score for confirming recognition result is more than confidence threshold value, just receive the recognition result,
Otherwise just refuse the recognition result.
However, the present inventor has found during the technical scheme in realizing the embodiment of the present invention, it is of the prior art
No matter when confidence threshold value all immobilizes, such as when environment is more noisy, because voice signal is made an uproar
The pollution of sound, causes the confidence score of recognition result to decline, so if judging according further to the high threshold of script, just
The ratio regular meeting that rate of failing to report, i.e. False Rejects can be increased increases；Or the ratio of current confidence threshold value setting is relatively low, and environment ratio
Quieter, the confidence score of recognition result is higher, if at this moment confirming according further to Low threshold, is increased by rate of false alarm, i.e., wrong
The ratio regular meeting that misconnection is received increases so that the hydraulic performance decline of speech recognition.
The content of the invention
The present invention provides the method for adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition, is used to solve existing skill
The confidence threshold value of speech recognition present in art immobilizes so that the relatively low technical problem of the performance of speech recognition.
One aspect of the present invention provides a kind of method of adjustment of confidence coefficient threshold of voice recognition, is applied to a support voice
In the electronic equipment of identification, the current confidence threshold value of the speech recognition is the first value, and methods described includes：Detect N number of ginseng
Number, obtains N number of testing result, wherein, N is the integer more than or equal to 1；At least based on an inspection in N number of testing result
Survey result and adjust the confidence threshold value so that the confidence threshold value is changed into second value from first value, wherein, described the
Two-value is the value identical or different with first value.
Preferably, the N number of parameter of detection, specially：Detect the ambient noise parameter residing for the electronic equipment；Detection
Operative scenario complex parameters residing for the electronic equipment；And/or the length parameter of the sentence to be confirmed after detection speech recognition.
It is preferably, described that the confidence threshold value is at least adjusted based on a testing result in N number of testing result,
Specifically include：Corresponding relation based on the ambient noise parameter, ambient noise parameter and confidence threshold value is adjusted；It is based on
The corresponding relation of the operative scenario complex parameters, operative scenario complex parameters and confidence threshold value is adjusted；And/or be based on
The corresponding relation of the length parameter, length parameter and confidence threshold value is adjusted.
Preferably, before the N number of parameter of detection, methods described also includes：Receive the first phonetic entry；Identification is described
First phonetic entry, obtains the first recognition result.
The second value is preferably based on to judge whether to receive first recognition result.
One embodiment of the invention also provides a kind of electronic equipment, supports speech recognition, the current confidence of the speech recognition
Degree threshold value is the first value, and the electronic equipment includes：Circuit board；Detection chip, is electrically connected at the circuit board, for detecting
N number of parameter, obtains N number of testing result, wherein, N is the integer more than or equal to 1；Process chip, is arranged on the circuit board,
A testing result for being at least based in N number of testing result adjusts the confidence threshold value so that the confidence level
Threshold value is changed into second value from first value, wherein, the second value is the value identical or different with first value.
Preferably, the detection chip is specifically for detecting the ambient noise parameter residing for the electronic equipment；Detection institute
State the operative scenario complex parameters residing for electronic equipment；And/or the length parameter of the sentence to be confirmed after detection speech recognition.
Preferably, the process chip is specifically for based on the ambient noise parameter, ambient noise parameter and confidence level
The corresponding relation of threshold value adjusts the confidence threshold value；Based on the operative scenario complex parameters, operative scenario complex parameters and
The corresponding relation of confidence threshold value adjusts the confidence threshold value；And/or based on the length parameter, length parameter and confidence level
The corresponding relation of threshold value adjusts the confidence threshold value.
Preferably, the electronic equipment also includes：One sound collection unit, for detecting N number of ginseng in the detection chip
Before number, the first phonetic entry is received；Voice recognition chip, for recognizing first phonetic entry, obtains the first identification knot
Really.
Preferably, the voice recognition chip is specifically additionally operable to judge whether that receiving described first knows based on the second value
Other result.
One or more technical schemes provided in the embodiment of the present invention, at least have the following technical effect that or advantage：
One embodiment of the invention is using one or more parameters of real-time detection（Such as environment, scene, sentence are in itself）, obtain
One or more testing results, then one or more testing results in these results confidence threshold value is adjusted
It is whole so that confidence threshold value can be changed into second value from the first value, wherein, second value is the value identical or different with the first value.
Consequently, it is possible to confidence threshold value can according to different environment, scene or different sentences be changed into adapt to environment, scene or
The confidence threshold value of sentence, so that phonetic recognization rate is higher, the performance of speech recognition is more preferable.
Further, also judge whether to receive recognition result based on adjusted confidence threshold value in one embodiment of the invention,
I.e. first real-time adjustment confidence threshold value, then carries out whether the credible of recognition result judged according to the confidence threshold value after adjustment,
So the judgement to recognition result is more reasonable, more accurately.
Brief description of the drawings
Fig. 1 is the method flow diagram of the control electronics in one embodiment of the invention；
Fig. 2 is the functional block diagram of the electronic equipment in one embodiment of the invention.
Specific embodiment
The present invention provides the method for adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition, is used to solve existing skill
The confidence threshold value of speech recognition present in art immobilizes so that the relatively low technical problem of the performance of speech recognition.
Technical scheme in the embodiment of the present invention is to solve above-mentioned technical problem, and general thought is as follows：
By one or more parameters of real-time detection（Such as environment, scene, sentence are in itself）, obtain one or more detections
As a result, one or more testing results then in these results are adjusted to confidence threshold value so that confidence level threshold
Value can be changed into second value from the first value, wherein, second value is the value identical or different with the first value.Consequently, it is possible to confidence level
Threshold value can be changed into adapting to the confidence level threshold of environment, scene or sentence according to different environment, scene or different sentences
Value, so that phonetic recognization rate is higher, the performance of speech recognition is more preferable.
In order to be better understood from above-mentioned technical proposal, below in conjunction with Figure of description and specific embodiment to upper
Technical scheme is stated to be described in detail.
One embodiment of the invention provides a kind of method of adjustment of confidence coefficient threshold of voice recognition, is applied to a support voice
On the electronic equipment of identification, the electronic equipment is, for example, the electronic equipments such as mobile phone, panel computer, notebook computer, and the voice is known
Not current confidence threshold value is the first value, for example, 80.
Fig. 1 is refer to, the method includes：
Step 101：N number of parameter is detected, N number of testing result is obtained, wherein, N is the integer more than or equal to 1；
Step 102：At least based on the testing result adjustment confidence threshold value in N number of testing result so that confidence level
Threshold value is changed into second value from the first value, wherein, second value is the value identical or different with the first value.
Wherein, in a step 101, N number of parameter is detected, in specific implementation process, the classification of N number of parameter needs inspection
Which parameter surveyed, can user set in advance by a user interface, for example on confidence score influence than larger
The length parameter of ambient noise parameter, operative scenario complex parameters, and/or sentence to be confirmed.In other embodiments, it is also possible to
It is the parameter of other influences confidence score, those skilled in the art can be configured according to actual needs.
In the present embodiment, N number of parameter is detected, can is specifically to detect the ambient noise parameter residing for electronic equipment；Inspection
Survey the operative scenario complex parameters residing for electronic equipment；And/or the length parameter of the sentence to be confirmed surveyed after speech recognition.
Only to detect one of parameter, it is also possible to detect above-mentioned all parameters.
Wherein, ambient noise parameter can be the decibel of noise, or when a phonetic entry is received, detect the voice
The signal to noise ratio of input, when signal to noise ratio or decibel than it is larger when, illustrate to the recognition result of the phonetic entry branch is lower,
So at this moment needing to detect ambient noise parameter.
And operative scenario complex parameters can be that each operative scenario complexity of electronic equipment is estimated in advance, pin
To each operative scenario, a complexity coefficient is all corresponded to, the complexity coefficient can be used as complex parameters, and electronic equipment may determine that
Which operative scenario oneself is currently at, and then obtains the complex parameters under the operative scenario.
And the length parameter of sentence to be confirmed, it is also possible to it is trained in advance, that is, trains the language of each sentence in vocabulary
Sentence length value, the number of the phoneme that can be included according to sentence determines, or according to comprising the number of Chinese character determine.
Further, the corresponding relation between N number of parameter and confidence threshold value is also trained, in the present embodiment, is continued with N number of
Parameter be above three parameter as a example by illustrate.
First, when ambient noise parameter is decibel or signal to noise ratio, for noise decibel, it may be determined that one
Individual decibel range out, the decibel range can be the electronic equipment may most quiet environment decibel to most noisy ring
The environment decibel, or other decibel ranges in border, then make identical in this decibel range for each decibel
Or different speech recognition trainings, the confidence value of the recognition result of the speech recognition under calculating correspondence decibel, can finally obtain
A series of confidence values under this decibel, then can take a series of average value of this confidence value, used as correspondence decibel
Confidence threshold value, may eventually form the mapping table of a decibel and confidence value.Certainly, in practical application, can also
Take a series of minimum of this confidence value, it is also possible to take a series of peak of this confidence value, or centre can be taken
Certain value confidence value, and cause that the ratio of the confidence value greater than or equal to the confidence value reaches 80% or other ratios,
For example, a series of confidence value that training is obtained is 80,81,82,78,79, at this moment that take 79, so according to statistical analysis
From the point of view of, if confidence threshold value is set into 79, then under the decibel, it will the recognition result for having 80% can all be received.
Further, it is possible to be further analyzed with confidence value mapping table to the decibel, a decibel can be obtained
Functional relation between confidence threshold value, naturally it is also possible to directly repeatedly trained to decibel and confidence value, obtains
Functional relation between decibel and confidence threshold value.
It is similar, when ambient noise parameter is signal to noise ratio, can train 0 to 1 signal to noise ratio scope under put
Confidence threshold, each signal to noise ratio in the range of, can all obtain a series of confidence value, then can take a series of this confidence
The average value of angle value, as the confidence threshold value of correspondence signal to noise ratio, may eventually form a decibel corresponding with confidence threshold value
Relation table.It is of course also possible to such as the situation of decibel, the distribution situation according to the confidence value under the signal to noise ratio takes other confidences
Degree threshold value.
Further, the functional relation between signal to noise ratio and confidence threshold value can also equally be obtained.
In addition, ambient noise becomes big, i.e., decibel becomes big, and signal to noise ratio declines, and causes the confidence score of normal voice to decline,
So needing to lower confidence threshold value, it is to avoid False Rejects increase, thus the result of last training can be confidence threshold value with
Increase, the decline of signal to noise ratio of decibel and decline, and then False Rejects can be reduced, and the amount specifically lowered can be by above-mentioned
Training method obtain.
Can also be according to above-mentioned training method for operative scenario complex parameters and the corresponding confidence threshold value of length parameter
It is trained, the length parameter of such as sentence, carries out speech recognition for the multiple sentences with same length parameter, and calculate
The confidence value of recognition result, obtains the confidence value distribution situation under the length parameter, and then can obtain corresponding confidence
The functional relation of degree threshold value or length parameter and confidence threshold value.
In addition, the change of operative scenario, such as, into a scene for complexity, the confidence score of speech recognition may
Uprise and be likely to meeting step-down, so confidence threshold value may be heightened to be likely to turn down, and the amount of adjustment can be by upper
The training method stated is obtained so that confidence threshold value and operative scenario are matched, that is, cause that False Rejects and mistake receive reduction.It is right
For the change of the length of sentence to be confirmed, the change situation with operative scenario is similar, will not be repeated here.
The various training methods of above-mentioned introduction, both can be manufacturer trains before dispatching from the factory models, or third party manufacturer
The model, or electronic equipment for training according to particular condition in use, for example in voice know every time by the process gradually trained
When other, electronic equipment can detect N number of parameter, and record N number of testing result, then again corresponding record in these testing results
Under, then the score of the confidence level of voice identification result will obtain a series of confidence values, that is, obtain a confidence value point
Cloth, electronic equipment carries out statistical analysis or calculating, for example according to the average value for taking a series of this confidence value principle or
Take centre certain value confidence value, and cause greater than or equal to the confidence value confidence value ratio reach 80% or
The principle of other ratios, or other principles, the automatic mapping table or letter for setting up simultaneously undated parameter and confidence threshold value
Number relational expression.Consequently, it is possible to electronic equipment can automatically update mapping table or functional relation by study, training,
So that the confidence threshold value after adjustment more conforms to actual conditions, so as to also cause that the performance of speech recognition is improved.
It is above-mentioned to describe the respective training method of different parameters respectively, but different parametric synthesis can also be got up
Train together, finally obtain the corresponding relation of multiple parameters and confidence threshold value, including corresponding table or functional relation.Enter one
Step, can also be trained for other specification according to above-mentioned training method, succinct for specification, will not be repeated here.
Certainly, above-mentioned training method is only citing, is not intended to limit the present invention, in specific implementation process, this area
Technology can also be trained using other training methods.
In a step 102, confidence threshold value is at least adjusted based on a testing result in N number of testing result, it is real one
Apply in example, step S102 can specifically include：
Based at least one of N number of testing result testing result, inquire about in N number of parameter the parameter of at least one parameter and
The mapping table of confidence threshold value, wherein, at least one of at least one parameter and N number of testing result inspection in N number of parameter
Result is surveyed to correspond to respectively；
Confidence level corresponding with least one testing result is determined in the mapping table of parameter and confidence threshold value
Threshold value；
Confidence threshold value corresponding with least one testing result is set to the confidence threshold value of speech recognition.
Specifically, it is exactly that the mapping table described in previous examples is directly inquired about according to testing result, because often
Individual parameter or multiple parameters one confidence threshold value of correspondence, as long as the confidence threshold value in the mapping table is set into voice knowing
Other confidence threshold value.Confidence threshold value after adjustment may with adjustment before it is identical, it is also possible to differ.
In another embodiment, step S102 can specifically include：
Based at least one of N number of testing result testing result, obtain at least one parameter in N number of parameter parameter and
Functional relation between confidence threshold value, wherein, in N number of parameter at least one parameter and N number of testing result at least one
Individual testing result is corresponded to respectively；
At least one testing result is substituted into the functional relation to be calculated, one is obtained and is calculated confidence threshold value；
The calculating confidence threshold value is set to the confidence threshold value of speech recognition.
Specifically, it is, by the functional relation described in examples detailed above, to calculate that the testing result is corresponding to put
Confidence threshold, then the confidence threshold value is set to the confidence threshold value of speech recognition.Confidence threshold value after adjustment
May with adjustment before it is identical, it is also possible to differ.
In specific implementation process, continue to continue to use above example, can be specifically：Based on ambient noise parameter, environment
The corresponding relation of noise parameter and confidence threshold value is adjusted；Based on operative scenario complex parameters, operative scenario complex parameters
Corresponding relation with confidence threshold value is adjusted；And/or the correspondence pass based on length parameter, length parameter and confidence threshold value
System is adjusted.Only can be adjusted according to one of parameter, it is also possible to be adjusted according to all parameters.
Wherein, for ambient parameter（Parameter is the situation of decibel）For operative scenario complex parameters, because can be
It is known a priori by, it is possible to adjust confidence threshold value previously according to ambient parameter and operative scenario complex parameters, will puts
Confidence threshold is adjusted to be adapted with the environment and operative scenario residing for electronic equipment so that the rate of false alarm of speech recognition and fail to report
Rate all than relatively low, i.e., mistake receive and False Rejects the ratio that occurs of situation all than relatively low.
And for ambient parameter（Parameter is the situation of the signal to noise ratio of sentence to be confirmed）With the length parameter of sentence to be confirmed
For because being after analyzing after sentence to be confirmed or recognizing sentence to be confirmed, further according to signal to noise ratio parameter or
Length parameter adjusts confidence threshold value.
Certainly, in practice, or after a phonetic entry is received, it is identified to the phonetic entry
Afterwards, confidence threshold value is adjusted further according to all parameters.
Further, it is possible to according to the confidence threshold value after adjustment, confirm recognition result, determine whether to receive to be somebody's turn to do
Recognition result, because the foundation for judging is according to the confidence threshold value after various parameters adjustment, so that rate of failing to report and wrong report
Rate is all reduced, so improve the performance of speech recognition.
Specifically, for example, ambient parameter is characterized by decibel, and current decibels are 90, represent that electronics sets
It is standby to be currently in a noisy environment, such as on road, represent that voice messaging is more serious by the ratio of noise pollution, and then lead
Cause the confidence score of speech recognition can be than relatively low, for example, at this moment user input one voice messaging " I is Xiao Ming ", passes through
Speech recognition, the recognition result for obtaining also " I is Xiao Ming ", but in order to further confirm that whether the recognition result may be used
Letter, then calculate the confidence score of the recognition result, and the confidence score of recognition result is learnt after calculating for 60 points, if
According to the method for prior art, if confidence threshold value is fixed on 80 points all the time, then by the confidence score 60 of recognition result
Divide and be compared with confidence threshold value 80, find the confidence threshold value of the confidence score less than default of recognition result, institute
It is insincere to judge the result, so the voice messaging would not be processed further, for example, other electric terminal is sent to, or
Show on the display unit, but actually this recognition result is believable, but because confidence threshold value sets too high, and is caused
The False Rejects recognition results.
However, by the confidence threshold value method of adjustment described in the present embodiment, electronic equipment by detecting ambient parameter,
According to the threshold value that automatically be adjusted to for confidence threshold value to be adapted with environment by ambient parameter, for example, it is 90 to detect decibels,
The mode that can be brought into functional relation by way of foregoing tabling look-up or by parameter, obtains a rational confidence level threshold
Value, e.g. 59, then 60 points of the confidence score of recognition result is compared for 59 points with confidence threshold value, result is identification
The confidence score of result is more than confidence threshold value, so illustrate that the recognition result is believable, so will believe the voice
Breath is for further processing.
Found out by above instantiation, the confidence after the method for adjustment of the confidence threshold value in the present embodiment is adjusted
Degree threshold value is more reasonable, and the ratio of False Rejects is reduced in speech recognition, that is, reduce rate of failing to report.Same reason, also may be used
To reduce the ratio that mistake receives, rate of false alarm is reduced.
Therefore, the confidence threshold value in the present embodiment can according to environment, operative scenario, the length of sentence to be confirmed change
Change and self-adaptative adjustment, also include that other are influenceed than larger other specification on confidence score certainly, so that confidence level
Threshold value can be adjusted in a rational value, reduced mistake and received and False Rejects so that the accuracy rate of speech recognition is higher,
Speech recognition performance is more preferable.
A kind of electronic equipment is also provided in one embodiment of the invention, the electronic equipment is, for example, mobile phone, panel computer, notes
The electronic equipments such as this computer, the electronic equipment supports speech recognition, and the current confidence threshold value of speech recognition is the first value.
As shown in Fig. 2 the electronic equipment includes：Circuit board 201；Detection chip 202, is electrically connected at circuit board 201, uses
In N number of parameter is detected, N number of testing result is obtained, wherein, N is the integer more than or equal to 1；Process chip 203, is arranged on circuit
On plate 201, a testing result for being at least based in N number of testing result adjusts confidence threshold value so that confidence threshold value
It is changed into second value from the first value, wherein, second value is the value identical or different with the first value.
Wherein, detection chip 202 is, for example, decibel instrument, the ambient noise for detecting electronic equipment output, decibel, or
It is to include：The sub- chip of sub- chip, Fourier transformation is backed up, sub- chip, the sub- chip of Fourier inversion is filtered, is calculated sub- chip
Chip, specific detection process is：By the voice signal of microphone typing first, the sub- chip of backup is then first passed through by the first voice
Signal is backed up, and generates one first backup voice signal；Then the first voice signal is changed by the sub- chip of Fourier transformation,
Filtered by filtering sub- chip in frequency domain, remove noise, the data eliminated after noise are then carried out into the sub- core of Fourier inversion
The treatment of piece, is then calculated the first voice signal of the first backup voice signal and removal noise in the sub- chip of calculating,
Draw ambient parameter signal to noise ratio.
In another embodiment, detection chip 202 can be with the operative scenario residing for direct detection electronic equipment.
In another embodiment, detection chip 202 can also be pronounciation processing chip, and then detect the length of sentence to be confirmed
Degree parameter.
In another embodiment, detection chip 202 includes kind described above chip, can detect described above each
Plant parameter.
Further, process chip 203 is specifically for based on ambient noise parameter, ambient noise parameter and confidence threshold value
Corresponding relation adjusts confidence threshold value；Based on the right of operative scenario complex parameters, operative scenario complex parameters and confidence threshold value
Adjustment confidence threshold value should be related to；And/or the corresponding relation adjustment confidence based on length parameter, length parameter and confidence threshold value
Degree threshold value.
Further, process chip 303 can be single process chip, it is also possible to be integrated in the centre of electronic equipment
In reason device.
In one embodiment, electronic equipment also includes：One sound collection unit, for detecting N number of ginseng in detection chip 202
Before number, the first phonetic entry is received；Voice recognition chip, for recognizing the first phonetic entry, obtains the first recognition result.Sound
Sound collecting unit, such as microphone；Voice recognition chip can be identical chip, or difference with process chip 203
Chip.
Further, voice recognition chip is specifically additionally operable to judge whether to receive the first recognition result based on second value.
Various embodiments above can individually be implemented, it is also possible to reference to implementation, and technical staff can be selected according to actual needs
Select.
Various change mode in confidence threshold value method of adjustment and instantiation in the embodiment of earlier figures 1 is equally applicable
In the electronic equipment of the present embodiment, by the foregoing detailed description to confidence threshold value method of adjustment, those skilled in the art can
To be apparent from the implementation of electronic equipment in the present embodiment, thus it is succinct for specification, will not be described in detail herein.
One or more technical schemes provided in the embodiment of the present invention, at least have the following technical effect that or advantage：
One embodiment of the invention is using one or more parameters of real-time detection（Such as environment, scene, sentence are in itself）, obtain
One or more testing results, then one or more testing results in these results confidence threshold value is adjusted
It is whole so that confidence threshold value can be changed into second value from the first value, wherein, second value is the value identical or different with the first value.
Consequently, it is possible to confidence threshold value can according to different environment, scene or different sentences be changed into adapt to environment, scene or
The confidence threshold value of sentence, so that phonetic recognization rate is higher, the performance of speech recognition is more preferable.
Further, also judge whether to receive recognition result based on adjusted confidence threshold value in one embodiment of the invention,
I.e. first real-time adjustment confidence threshold value, then carries out whether the credible of recognition result judged according to the confidence threshold value after adjustment,
So the judgement to recognition result is more reasonable, more accurately.
It should be understood by those skilled in the art that, embodiments of the invention can be provided as method, system or computer program
Product.Therefore, the present invention can be using the reality in terms of complete hardware embodiment, complete software embodiment or combination software and hardware
Apply the form of example.And, the present invention can be used and wherein include the computer of computer usable program code at one or more
Usable storage medium（Including but not limited to magnetic disk storage and optical memory etc.）The shape of the computer program product of upper implementation
Formula.
The present invention is with reference to method according to embodiments of the present invention, equipment（System）And the flow of computer program product
Figure and/or block diagram are described.It should be understood that every first-class during flow chart and/or block diagram can be realized by computer program instructions
The combination of flow and/or square frame in journey and/or square frame and flow chart and/or block diagram.These computer programs can be provided
The processor of all-purpose computer, special-purpose computer, Embedded Processor or other programmable data processing devices is instructed to produce
A raw machine so that produced for reality by the instruction of computer or the computing device of other programmable data processing devices
The device of the function of being specified in present one flow of flow chart or multiple one square frame of flow and/or block diagram or multiple square frames.
These computer program instructions may be alternatively stored in can guide computer or other programmable data processing devices with spy
In determining the computer-readable memory that mode works so that instruction of the storage in the computer-readable memory is produced and include finger
Make the manufacture of device, the command device realize in one flow of flow chart or multiple one square frame of flow and/or block diagram or
The function of being specified in multiple square frames.
These computer program instructions can be also loaded into computer or other programmable data processing devices so that in meter
Series of operation steps is performed on calculation machine or other programmable devices to produce computer implemented treatment, so as in computer or
The instruction performed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one
The step of function of being specified in individual square frame or multiple square frames.
Obviously, those skilled in the art can carry out various changes and modification without deviating from essence of the invention to the present invention
God and scope.So, if these modifications of the invention and modification belong to the scope of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to comprising these changes and modification.
Claims (10)
1. a kind of method of adjustment of confidence coefficient threshold of voice recognition, is applied in the electronic equipment of a support speech recognition, institute
The current confidence threshold value of speech recognition is stated for the first value, it is characterised in that methods described includes：
N number of parameter is detected, N number of testing result is obtained, wherein, N is the integer more than or equal to 1；
The confidence threshold value is at least adjusted based on a testing result in N number of testing result so that the confidence level
Threshold value is changed into second value from first value, wherein, the second value is the value identical or different with first value；
The confidence score of voice identification result is compared with the second value, and is determined whether to receive according to comparative result
Institute's speech recognition result.
2. the method for claim 1, it is characterised in that the N number of parameter of detection, specially：
Detect the ambient noise parameter residing for the electronic equipment；
Detect the operative scenario complex parameters residing for the electronic equipment；And/or
The length parameter of the sentence to be confirmed after detection speech recognition.
3. method as claimed in claim 2, it is characterised in that described at least based on an inspection in N number of testing result
Survey result and adjust the confidence threshold value, specifically include：
Corresponding relation based on the ambient noise parameter, ambient noise parameter and confidence threshold value is adjusted；
Corresponding relation based on the operative scenario complex parameters, operative scenario complex parameters and confidence threshold value is adjusted；
And/or
Corresponding relation based on the length parameter, length parameter and confidence threshold value is adjusted.
4. the method for claim 1, it is characterised in that before the N number of parameter of detection, methods described also includes：
Receive the first phonetic entry；
First phonetic entry is recognized, the first recognition result is obtained.
5. method as claimed in claim 4, it is characterised in that judge whether to receive first identification based on the second value
As a result.
6. a kind of electronic equipment, supports speech recognition, and the current confidence threshold value of the speech recognition is the first value, and its feature exists
In the electronic equipment includes：
Circuit board；
Detection chip, is electrically connected at the circuit board, for detecting N number of parameter, obtains N number of testing result, wherein, N is big
In the integer equal to 1；
Process chip, is arranged on the circuit board, and a testing result for being at least based in N number of testing result is adjusted
The whole confidence threshold value so that the confidence threshold value is changed into second value from first value, wherein, the second value be with
The identical or different value of first value；The confidence score of voice identification result is compared with the second value, and root
Determine whether to receive institute's speech recognition result according to comparative result.
7. electronic equipment as claimed in claim 6, it is characterised in that the detection chip sets specifically for detecting the electronics
Standby residing ambient noise parameter；Detect the operative scenario complex parameters residing for the electronic equipment；And/or detection speech recognition
The length parameter of sentence to be confirmed afterwards.
8. electronic equipment as claimed in claim 7, it is characterised in that the process chip based on the environment specifically for being made an uproar
The corresponding relation of sound parameter, ambient noise parameter and confidence threshold value adjusts the confidence threshold value；Based on the operative scenario
The corresponding relation of complex parameters, operative scenario complex parameters and confidence threshold value adjusts the confidence threshold value；And/or based on institute
The corresponding relation for stating length parameter, length parameter and confidence threshold value adjusts the confidence threshold value.
9. electronic equipment as claimed in claim 6, it is characterised in that the electronic equipment also includes：
One sound collection unit, for before the detection chip detects N number of parameter, receiving the first phonetic entry；
Voice recognition chip, for recognizing first phonetic entry, obtains the first recognition result.
10. electronic equipment as claimed in claim 9, it is characterised in that the voice recognition chip is specifically additionally operable to based on institute
Second value is stated to judge whether to receive first recognition result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210272154.7A CN103578468B (en) | 2012-08-01 | 2012-08-01 | The method of adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210272154.7A CN103578468B (en) | 2012-08-01 | 2012-08-01 | The method of adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103578468A CN103578468A (en) | 2014-02-12 |
| CN103578468B true CN103578468B (en) | 2017-06-27 |
Family
ID=50050121
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210272154.7A Active CN103578468B (en) | 2012-08-01 | 2012-08-01 | The method of adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103578468B (en) |
Families Citing this family (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103065631B (en) * | 2013-01-24 | 2015-07-29 | 华为终端有限公司 | A kind of method of speech recognition, device |
| CN103971680B (en) | 2013-01-24 | 2018-06-05 | 华为终端（东莞）有限公司 | A kind of method, apparatus of speech recognition |
| US9335703B1 (en) * | 2014-10-16 | 2016-05-10 | Kodak Alaris Inc. | Audio detection of medium jam |
| US9685156B2 (en) * | 2015-03-12 | 2017-06-20 | Sony Mobile Communications Inc. | Low-power voice command detector |
| KR102420450B1 (en) * | 2015-09-23 | 2022-07-14 | 삼성전자주식회사 | Voice Recognition Apparatus, Voice Recognition Method of User Device and Computer Readable Recording Medium |
| CN106653010B (en) * | 2015-11-03 | 2020-07-24 | 络达科技股份有限公司 | Electronic device and method for waking up through voice recognition |
| CN105702253A (en) * | 2016-01-07 | 2016-06-22 | 北京云知声信息技术有限公司 | Voice awakening method and device |
| CN106338924A (en) * | 2016-09-23 | 2017-01-18 | 广州视源电子科技股份有限公司 | Method and device for automatically adjusting equipment operation parameter threshold |
| CN106384594A (en) * | 2016-11-04 | 2017-02-08 | 湖南海翼电子商务股份有限公司 | On-vehicle terminal for voice recognition and method thereof |
| CN108550365B (en) * | 2018-02-01 | 2021-04-02 | 云知声智能科技股份有限公司 | Threshold value self-adaptive adjusting method for off-line voice recognition |
| CN108847219B (en) * | 2018-05-25 | 2020-12-25 | 台州智奥通信设备有限公司 | Awakening word preset confidence threshold adjusting method and system |
| CN109003607B (en) * | 2018-07-12 | 2021-06-01 | Oppo广东移动通信有限公司 | Voice recognition method, voice recognition device, storage medium and electronic equipment |
| CN110197663B (en) * | 2019-06-30 | 2022-05-31 | 联想(北京)有限公司 | Control method and device and electronic equipment |
| CN110580897B (en) * | 2019-08-23 | 2022-07-12 | Oppo广东移动通信有限公司 | Audio verification method and device, storage medium and electronic equipment |
| CN110544468B (en) * | 2019-08-23 | 2022-07-12 | Oppo广东移动通信有限公司 | Application awakening method and device, storage medium and electronic equipment |
| CN110602391B (en) * | 2019-08-30 | 2021-08-24 | Oppo广东移动通信有限公司 | Photographing control method, device, storage medium and electronic device |
| CN110880318B (en) * | 2019-11-27 | 2023-04-18 | 云知声智能科技股份有限公司 | Voice recognition method and device |
| CN111161012B (en) * | 2019-12-05 | 2020-10-02 | 广州二空间信息服务有限公司 | Information pushing method and device and computer equipment |
| CN111326148B (en) * | 2020-01-19 | 2021-02-23 | 北京世纪好未来教育科技有限公司 | Confidence correction and model training method, device, equipment and storage medium thereof |
| CN111739542B (en) * | 2020-05-13 | 2023-05-09 | 深圳市微纳感知计算技术有限公司 | Method, device and equipment for detecting characteristic sound |
| US11769484B2 (en) * | 2020-09-11 | 2023-09-26 | International Business Machines Corporation | Chaos testing for voice enabled devices |
| CN112434681B (en) * | 2021-01-27 | 2021-05-07 | 武汉星巡智能科技有限公司 | Smart camera self-training confidence threshold selection method, device and equipment |
| CN112802483B (en) * | 2021-04-14 | 2021-06-29 | 南京山猫齐动信息技术有限公司 | Method, device and storage medium for optimizing intention recognition confidence threshold |
| CN113330513B (en) * | 2021-04-20 | 2024-08-27 | 华为技术有限公司 | Voice information processing method and equipment |
| CN113823276B (en) * | 2021-09-29 | 2023-06-02 | 成都启英泰伦科技有限公司 | Voice recognition threshold setting method |
| CN114743571A (en) * | 2022-04-08 | 2022-07-12 | 北京字节跳动网络技术有限公司 | Audio processing method and device, storage medium and electronic equipment |
| CN115294974B (en) * | 2022-06-28 | 2025-02-28 | 京东科技信息技术有限公司 | A speech recognition method, device, equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1264888A (en) * | 1998-12-17 | 2000-08-30 | 索尼国际(欧洲)股份有限公司 | Semi-monitoring speaker self-adaption |
| CN1434406A (en) * | 2001-12-12 | 2003-08-06 | 微软公司 | System and method for self-adaptive communication prioritization and route |
| CN1842842A (en) * | 2003-08-29 | 2006-10-04 | 松下电器产业株式会社 | A method and device for improving speech recognition based on auxiliary information |
| CN101206857A (en) * | 2006-12-19 | 2008-06-25 | 国际商业机器公司 | Method and system for modifying speech processing arrangement |
| CN101272416A (en) * | 2006-11-28 | 2008-09-24 | 通用汽车公司 | Voice Dialing Using Deny References |
| CN101447185A (en) * | 2008-12-08 | 2009-06-03 | 深圳市北科瑞声科技有限公司 | Audio frequency rapid classification method based on content |
| US7657433B1 (en) * | 2006-09-08 | 2010-02-02 | Tellme Networks, Inc. | Speech recognition accuracy with multi-confidence thresholds |
| CN101640042A (en) * | 2008-07-29 | 2010-02-03 | 佳能株式会社 | Information processing method and information processing apparatus |
| CN102136269A (en) * | 2010-01-22 | 2011-07-27 | 微软公司 | Speech recognition analysis via identification information |
| CN102194454A (en) * | 2010-03-05 | 2011-09-21 | 富士通株式会社 | Equipment and method for detecting key word in continuous speech |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8005668B2 (en) * | 2004-09-22 | 2011-08-23 | General Motors Llc | Adaptive confidence thresholds in telematics system speech recognition |
| US8396715B2 (en) * | 2005-06-28 | 2013-03-12 | Microsoft Corporation | Confidence threshold tuning |
-
2012
- 2012-08-01 CN CN201210272154.7A patent/CN103578468B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1264888A (en) * | 1998-12-17 | 2000-08-30 | 索尼国际(欧洲)股份有限公司 | Semi-monitoring speaker self-adaption |
| CN1434406A (en) * | 2001-12-12 | 2003-08-06 | 微软公司 | System and method for self-adaptive communication prioritization and route |
| CN1842842A (en) * | 2003-08-29 | 2006-10-04 | 松下电器产业株式会社 | A method and device for improving speech recognition based on auxiliary information |
| US7657433B1 (en) * | 2006-09-08 | 2010-02-02 | Tellme Networks, Inc. | Speech recognition accuracy with multi-confidence thresholds |
| CN101272416A (en) * | 2006-11-28 | 2008-09-24 | 通用汽车公司 | Voice Dialing Using Deny References |
| CN101206857A (en) * | 2006-12-19 | 2008-06-25 | 国际商业机器公司 | Method and system for modifying speech processing arrangement |
| CN101640042A (en) * | 2008-07-29 | 2010-02-03 | 佳能株式会社 | Information processing method and information processing apparatus |
| CN101447185A (en) * | 2008-12-08 | 2009-06-03 | 深圳市北科瑞声科技有限公司 | Audio frequency rapid classification method based on content |
| CN102136269A (en) * | 2010-01-22 | 2011-07-27 | 微软公司 | Speech recognition analysis via identification information |
| CN102194454A (en) * | 2010-03-05 | 2011-09-21 | 富士通株式会社 | Equipment and method for detecting key word in continuous speech |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103578468A (en) | 2014-02-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103578468B (en) | The method of adjustment and electronic equipment of a kind of confidence coefficient threshold of voice recognition | |
| US12380895B2 (en) | Analysing speech signals | |
| US8762144B2 (en) | Method and apparatus for voice activity detection | |
| US20180374487A1 (en) | Detection of replay attack | |
| EP1973104B1 (en) | Method and apparatus for estimating noise by using harmonics of a voice signal | |
| CN101510423B (en) | Multilevel interactive pronunciation quality estimation and diagnostic system | |
| CN110299142B (en) | Voiceprint recognition method and device based on network convergence | |
| CN111325386B (en) | Method, device, terminal and storage medium for predicting the running state of vehicles | |
| CN102446504B (en) | Speech/music recognition method and device | |
| CN103646649A (en) | High-efficiency voice detecting method | |
| CN109614881A (en) | Biometric authentication method, device and storage device capable of adaptively adjusting threshold | |
| CN110473552A (en) | Speech recognition authentication method and system | |
| US20060100866A1 (en) | Influencing automatic speech recognition signal-to-noise levels | |
| CN106782508A (en) | The cutting method of speech audio and the cutting device of speech audio | |
| CN103405217A (en) | System and method for multi-dimensional measurement of dysarthria based on real-time articulation modeling technology | |
| CN113077812A (en) | Speech signal generation model training method, echo cancellation method, device and equipment | |
| WO2018095167A1 (en) | Voiceprint identification method and voiceprint identification system | |
| EP4254409A1 (en) | Voice detection method | |
| CN105810222A (en) | Defect detection method, device and system for audio equipment | |
| CN107274892A (en) | Method for distinguishing speek person and device | |
| CN104282303A (en) | Method and electronic device for speech recognition using voiceprint recognition | |
| CN112216285B (en) | Multi-user session detection method, system, mobile terminal and storage medium | |
| CN114678040A (en) | Voice consistency detection method, device, equipment and storage medium | |
| CN109389993A (en) | A kind of data under voice method, apparatus, equipment and storage medium | |
| CN106340310B (en) | Speech detection method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |