CN109346099A - A kind of iterated denoising method and chip based on speech recognition - Google Patents
A kind of iterated denoising method and chip based on speech recognition Download PDFInfo
- Publication number
- CN109346099A CN109346099A CN201811512492.7A CN201811512492A CN109346099A CN 109346099 A CN109346099 A CN 109346099A CN 201811512492 A CN201811512492 A CN 201811512492A CN 109346099 A CN109346099 A CN 109346099A
- Authority
- CN
- China
- Prior art keywords
- voice signal
- noise
- confidence value
- acoustic frame
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 45
- 230000005236 sound signal Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 235000013399 edible fruits Nutrition 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 2
- 244000287680 Garcinia dulcis Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/20—Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
- G10L2021/02161—Number of inputs available containing the signal or the noise to be suppressed
- G10L2021/02166—Microphone arrays; Beamforming
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Manipulator (AREA)
Abstract
The present invention discloses a kind of iterated denoising method and chip based on speech recognition, comprising: step 1: determining targeted voice signal and its objective degrees of confidence value；Step 2: selection is worth matched noise data with objective degrees of confidence from noise database, and control noise data has acoustic frame to participate in predenoising processing with unlabelled in targeted voice signal；Step 3: judging whether the predenoising processing result is greater than the predetermined threshold, be to enter step 4, otherwise enter step 5；Step 4: the corresponding sound frame flag of the predenoising processing result is had into acoustic frame for what is denoised in the targeted voice signal；Step 5: judge the predenoising processing result confidence value and the objective degrees of confidence value absolute difference whether less than one confidence threshold value, be, by the corresponding sound frame flag of the predenoising processing result be the targeted voice signal in denoised have acoustic frame；Otherwise the objective degrees of confidence value is adjusted, the step 2 is returned again to.
Description
Technical field
The invention belongs to robotic technology field more particularly to a kind of iterated denoising methods and core based on speech recognition
Piece.
Background technique
Existing mobile robot (such as sweeping robot, window wiping robot, floor-mopping robot in clean robot etc.
Electromechanical integration equipment) noise that can generate during the work time, although although the voice pick device to circulate on the market can be with
Voice pickup is carried out to the voice signal that user issues, but while being picked up to the voice signal that user issues, is led to
Often also voice pickup can will be carried out to the noise that can be generated in the robot course of work, so that in the voice signal that equipment is picked up
It is mingled with a large amount of outside noise, corresponding speech discrimination accuracy is not high, this, which will seriously affect robot, (extraneous voice
Imitate signal) identification, and it is voice-based explanation and make logic decision (such as executing relevant path planning).
In the prior art, it is generally in the method that the front end of voice signal denoises, according to classification of speech signals as a result, choosing
Suitable voice signal is selected, and inhibits undesirable voice signal, but is complex to the method for classification of speech signals, no
Only noise reduction is not thorough, and audio identification efficiency is not high, always has remaining speech frame not processed, influences the effect of speech recognition
Fruit.
Summary of the invention
In order to overcome the above technical defects, the present invention proposes following technical scheme:
A kind of iterated denoising method based on speech recognition, which includes: step 1: being obtained from microphone array
Voice signal in determine targeted voice signal, correspondingly obtain objective degrees of confidence value；Step 2: being selected from noise database
It is worth matched noise data with objective degrees of confidence, and control noise data to have acoustic frame participation with unlabelled in targeted voice signal
Predenoising processing, to obtain predenoising processing result corresponding with noise data；Step 3: judging the predenoising processing result
Whether it is greater than the predetermined threshold, is to enter step 4, otherwise enters step 5；Step 4: by the predenoising processing result pair
The sound frame flag answered has acoustic frame for what is denoised in the targeted voice signal；Step 5: judging the predenoising processing result
Confidence value and the objective degrees of confidence value absolute difference whether less than one confidence threshold value, be pre- to be gone described
The corresponding sound frame flag of processing result of making an uproar has acoustic frame for what is denoised in the targeted voice signal；Otherwise the target is adjusted
Confidence value returns again to the step 2；Wherein, targeted voice signal has acoustic frame comprising associated with control instruction.It is described to change
For denoising method by being judged twice the predenoising processing result, comprehensively dispose each in the targeted voice signal
There is acoustic frame, be conducive to the completeness of denoising and improve the accuracy of speech de-noising.
Further, the step 1 specifically includes: language acquired in the microphone array is identified by speech engine
Sound signal has acoustic frame, and when the signal-to-noise ratio numerical value for having acoustic frame is greater than predetermined snr threshold, this is had the corresponding voice of acoustic frame
Signal is determined as the targeted voice signal, then from the target for having and extracting the corresponding targeted voice signal in acoustic frame
Confidence value, wherein described to have acoustic frame include confidence value and signal-to-noise ratio numerical value based on speech recognition signal.According to predetermined letter
It makes an uproar than threshold value and filters out targeted voice signal, identifying processing targetedly is carried out to specific voice signal, improve noise
The precision of speech recognition under environment.
Further, in the step 2, the selection from noise database is worth matched noise number with objective degrees of confidence
According to specifically including: judging in the noise database with the presence or absence of one small with the absolute difference of the objective degrees of confidence value
It is that then the determining default noise data is the described and mesh in the confidence value of the default noise data of predetermined noise threshold
Mark the matched noise data of confidence value.According to the real-time matching degree of the confidence value of noise signal and noise database, selection
Matched preset noise signal improves the accuracy of denoising operation.
Further, the method for the predenoising processing specifically includes: it first controls the noise data and carries out reverse phase processing,
Obtain a reverse phase noise signal；The reverse phase noise signal is controlled again carry out mixing with the targeted voice signal be superimposed, with
Obtain the predenoising processing result corresponding with the noise data.The predenoising processing method is simple and efficient.
Further, the method for the adjustment objective degrees of confidence value includes: according to unmarked in the targeted voice signal
The confidence value for having acoustic frame and the current objective degrees of confidence value difference, the current objective degrees of confidence value is tuned up
Or it turns down.Be conducive to subsequent based on the judgement screening for having acoustic frame unlabelled in the targeted voice signal, raising iterative processing
The accuracy of process.
A kind of chip, the chip is for storing the corresponding program code of the iterated denoising method.By selectively
There is acoustic frame to carry out denoising targeted voice signal, threshold value is intelligently set and marks what is currently denoised to have acoustic frame, thus
Inhibit influence of the noise to speech recognition, so that denoising effect is more thorough.After obtaining targeted voice signal, carry out first
Denoising, to improve recognition accuracy.And it according to the real-time matching degree of noise signal and noise database, is adjusted flexibly and sets
Certainty value is to improve denoising efficiency, to further increase audio identification efficiency.
Compared with prior art, technical solution of the present invention is carrying out noise suppression preprocessing after obtaining targeted voice signal
During, it is selectively right according to the confidence value real-time matching degree of the confidence value of noise signal and noise database
Targeted voice signal has acoustic frame to carry out denoising, and by the way that confidence value is adjusted flexibly, and combines of noise database
Denoising is carried out with data, to improve the denoising rate of the targeted voice signal.
Detailed description of the invention
Fig. 1 is a kind of process for iterated denoising method based on speech recognition that one embodiment of the present invention provides
Figure.
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 is retouched in detail
It states.
As shown in fig.1, the embodiment of the present invention provides a kind of iterated denoising method based on speech recognition, as described
A kind of embodiment of iterated denoising method, comprising:
Step S101, the voice signal that a particular orientation transmits is obtained from microphone array, is stored in advance based on speech engine
Database information domain analysis go determine targeted voice signal, with realize orientation voice pick up, reduce outside noise interference.So
After enter step S102.The targeted voice signal includes the voice data of the oral control command said of user or machine input,
Correspondingly, objective degrees of confidence value is obtained based on targeted voice signal, under the present embodiment, objective degrees of confidence value refers to the movement
Robot can be used to the credibility for indicating the preliminary recognition result of voice to the degree of the authenticity information of special sound signal
Numerical value judges the correctness of recognition result according to confidence threshold value, then presents to reduce erroneous judgement.The target language that user says
Sound signal is " returning seat charging ", then in the voice data identification process, the objective degrees of confidence value of return includes: sentence confidence level N.
Optionally, voice signal acquired in the microphone array is identified by the speech engine has acoustic frame,
Can include and control instruction phase by related voice Characteristics Detection algorithm, the targeted voice signal in the microphone array
It is associated to have acoustic frame, so that the targeted voice signal is convertible into multiple speech frames associated with user spoken utterances, wherein language
Sound frame may include having acoustic frame and silent frame, which can be executed by various known technologies.When the signal-to-noise ratio numerical value for having acoustic frame
When greater than predetermined snr threshold, there is the corresponding voice signal of acoustic frame to be determined as the targeted voice signal this, then from institute
State the objective degrees of confidence value that the corresponding targeted voice signal is extracted in acoustic frame, wherein described to have acoustic frame including based on language
The confidence value and signal-to-noise ratio numerical value of sound identification signal.
It should be noted that it is described have acoustic frame can be measured using signal-to-noise ratio it includes noise energy level, the letter
Make an uproar than be voice data Yu noise data power ratio, usually indicated with decibels, general signal-to-noise ratio is higher to show noise number
It is smaller according to power, otherwise on the contrary.The noise energy level is the size for reacting noise data energy in user voice data.
Signal-to-noise ratio and noise energy level combine, and indicate the noise size.
Step S102, from selection and the matched noise data of objective degrees of confidence value in noise database is pre-configured, so
After enter step S103.Specifically, according to the targeted voice signal include described in have corresponding noise data in acoustic frame, from
Search is based on a default noise data in the pre-configuration noise database, then judges that the objective degrees of confidence value is made an uproar with default
Whether the absolute difference of the confidence value of sound data is less than predetermined noise threshold, is that then the determining default noise data is institute
It states and the matched noise data of objective degrees of confidence value.Under the embodiment of the present invention, due to being produced in the working region of robot
Raw noise is more stable, therefore the pre-configuration noise database does not need just to carry out making poor ratio by real-time update noise data
Compared with compared with the existing technology, the software for reducing speech recognition is born, and can leave relevant software resource at subsequent denoising for
Reason.
Preferably, institute's noise data in the targeted voice signal and the pre-configuration noise database can be carried out
To obtain all voice similarity values, the weighted average for being then based on all voice similarity values determines described make a reservation for for comparison
Noise threshold.In addition, the highest result of discrimination can be selected as most by multiple noise databases, and from multiple databases
Matching result afterwards.To improve the discrimination of the work noise of robot.
Step S103, control the noise data has acoustic frame to participate in predenoising with unlabelled in the targeted voice signal
Processing, to obtain predenoising processing result corresponding with the noise data；Specifically, the method for the predenoising processing is specific
Include: first to control the noise data to carry out reverse phase processing, obtains a reverse phase noise signal；The reverse phase noise letter is controlled again
Number carrying out mixing with the targeted voice signal is superimposed, and is tied with obtaining the predenoising processing corresponding with the noise data
Fruit obtains predenoising treated voice messaging to eliminate the noise signal in the targeted voice signal.
Step S104, judge whether the predenoising processing result is greater than predetermined threshold, be to enter step S105, otherwise
Enter step S106；The predetermined threshold be it is pre-stored, can be used to the distortion level for measuring voice signal.If described pre-
Denoising result is greater than the predetermined threshold, then the predenoising processing result indicates having from the targeted voice signal
Undesirable noise is removed in acoustic frame, i.e. influence of these noises to recognition result has been eliminated.If the predenoising processing knot
Fruit is less than the predetermined threshold, then needs further to adjust denoising, each in the received targeted voice signal to guarantee
It is a to have acoustic frame that be handled, so that Speech Signal De-Noise is more thorough, and then improve the integrality of the targeted voice signal
With the accuracy rate of identification.
Step S105, it is greater than the judging result of the predetermined threshold according to the predenoising processing result, at predenoising
The corresponding sound frame flag of reason result is the acoustic frame that has denoised in the targeted voice signal, and the targeted voice signal
It is remaining it is unlabelled have acoustic frame may be unsatisfactory for after the predenoising processing predenoising processing result be greater than it is described predetermined
This condition of threshold value, needs to wait for the denoising of subsequent step, is then uniformly converted to phonetic control command, for controlling movement
Robot.
Step S106, the confidence value for judging the predenoising processing result and the difference of the objective degrees of confidence value are absolute
Value whether less than one confidence threshold value, be to enter step S107, otherwise enter step S108.The predenoising processing result
Confidence value be the predenoising processing result be less than the predetermined threshold under the premise of, for predenoising processing described in
The numerical value of the credibility of the recognition result of targeted voice signal, and the confidence threshold value can be used as the targeted voice signal
The evaluation index of correct recognition rata after being disturbed.By judge the predenoising processing result confidence value and the target
The absolute difference of confidence value whether less than one confidence threshold value, to be further processed the residue of the targeted voice signal
The unlabelled noise signal for having acoustic frame, to improve the comprehensive and accuracy of the targeted voice signal denoising.
It step S107, is to have been gone in the targeted voice signal by the corresponding sound frame flag of the predenoising processing result
That makes an uproar has acoustic frame, so that the unlabelled processing for having acoustic frame of step S105 judgement screening is realized, to improve the essence of speech recognition
Degree；And have acoustic frame there are still unlabelled in the targeted voice signal, at this time, it is meant that the current objective degrees of confidence value
Matched noise data is unobvious to the unlabelled predenoising effect for having acoustic frame, and error is larger.
Step S108, according to the confidence value for having acoustic frame unlabelled in the targeted voice signal and the current mesh
The size of the difference for marking confidence value, the current objective degrees of confidence value is tuned up or is turned down.Under the present embodiment, when the mesh
When the unlabelled confidence value for having acoustic frame is greater than the current objective degrees of confidence value in poster sound signal, described in current
Objective degrees of confidence value tunes up accordingly, on the contrary then turn down accordingly, then return step S102, selection and the mesh adjusted
The mark matched noise data of confidence value makees further denoising.This is apparently based on the current objective degrees of confidence value
Parameters revision process, be then based on that corrected parameter rejudges label denoising has acoustic frame, in this way, by successive ignition, circulation
Until all in the targeted voice signal have acoustic frame to complete denoising.According to the real-time of noise signal and noise database
Confidence value is adjusted flexibly to improve denoising efficiency in matching degree.Then there is acoustic frame for what is denoised in the targeted voice signal
It is converted to phonetic control command and goes control mobile robot.Since the targeted voice signal includes cyclic component, so,
There is the rule of periodic iterations during executing aforementioned denoising method, avoid correcting the objective degrees of confidence value at random, add
Fastly to the judgement speed of the targeted voice signal, to improve the working efficiency of denoising.
Technical solution of the present invention obtains the targeted voice signal that user issues under the noise scenarios that robot works, and
The empirical data of noise database according to the pre-stored data and the targeted voice signal, control the empirical data of noise database
Reverse phase handles the noise to inhibit the targeted voice signal；Simultaneously according to the real-time matching degree of noise signal and noise database
It is adjusted flexibly associated confidence value, then has acoustic frame by what judgement filtered out denoising, greatly improve the completeness of denoising, in turn
Phonetic recognization rate under noise circumstance.
A kind of chip, the chip is for storing the corresponding program code of the iterated denoising method.The chip is using dedicated
Integrated control chip, these chips can parse control instruction either internally or externally, and export corresponding control signal, with
The execution unit of control robot is acted accordingly.The built-in chip type is clear for controlling in the clean robot
Clean robot executes the iterated denoising method, and the targeted voice signal is converted to matched control by treated
Instruction, and corresponding operation is executed according to the control instruction.Wherein, the predenoising processing can be carried out using subtraction circuit
Signal subtraction can also be combined using phase inverter and add circuit and carry out signal subtraction, these circuits can be together with processor
It is integrated into dedicated processes chip, specifically corresponding configuration can be carried out according to design requirement.After having filtered out internal noise interference, place
Reason device again parses the signal after filtering out, and parses external sound signal, is converted to matched control instruction with reality
Now to the control of robot.How external sound signal is parsed as robot, belongs to the existing technology for having been able to realize,
Details are not described herein again.
Finally it should be noted that: the above embodiments are merely illustrative of the technical scheme of the present invention and are not intended to be limiting thereof；To the greatest extent
The present invention is described in detail with reference to preferred embodiments for pipe, it should be understood by those ordinary skilled in the art that: still
It can modify to a specific embodiment of the invention or some technical features can be equivalently replaced；Without departing from this hair
The spirit of bright technical solution should all cover within the scope of the technical scheme claimed by the invention.
Claims (6)
1. a kind of iterated denoising method based on speech recognition characterized by comprising
Step 1: determining targeted voice signal from the voice signal that microphone array obtains, correspondingly obtain objective degrees of confidence
Value；
Step 2: selection is worth matched noise data with objective degrees of confidence from noise database, and controls noise data and target
It is unlabelled in voice signal to there is acoustic frame to participate in predenoising processing, to obtain predenoising processing result corresponding with noise data；
Step 3: judging whether the predenoising processing result is greater than the predetermined threshold, be to enter step 4, otherwise enter step
Rapid 5；
Step 4: the corresponding sound frame flag of the predenoising processing result is had for what is denoised in the targeted voice signal
Acoustic frame；
Step 5: judge the predenoising processing result confidence value and the objective degrees of confidence value absolute difference whether
It is the targeted voice signal that less than one confidence threshold value, which is then by the corresponding sound frame flag of the predenoising processing result,
In denoised have acoustic frame；Otherwise the objective degrees of confidence value is adjusted, the step 2 is returned again to；Wherein, targeted voice signal packet
There is acoustic frame containing associated with control instruction.
2. iterated denoising method according to claim 1, which is characterized in that the step 1 specifically includes:
Voice signal acquired in the microphone array is identified by speech engine has acoustic frame, when the signal-to-noise ratio for having acoustic frame
When numerical value is greater than predetermined snr threshold, there is the corresponding voice signal of acoustic frame to be determined as the targeted voice signal this, then
From the objective degrees of confidence value for having and extracting the corresponding targeted voice signal in acoustic frame, wherein described to have acoustic frame include base
In the confidence value and signal-to-noise ratio numerical value of speech recognition signal.
3. iterated denoising method according to claim 1, which is characterized in that described from noise database in the step 2
Selection is worth matched noise data with objective degrees of confidence, specifically includes:
Judge to be less than predetermined make an uproar with the presence or absence of the absolute difference of one and the objective degrees of confidence value in the noise database
The confidence value of the default noise data of sound threshold value is that then the determining default noise data is the described and objective degrees of confidence
It is worth matched noise data.
4. iterated denoising method according to claim 1, which is characterized in that the method for the predenoising processing specifically includes:
It first controls the noise data and carries out reverse phase processing, obtain a reverse phase noise signal；
The reverse phase noise signal is controlled again carry out mixing with the targeted voice signal be superimposed, it is described with the noise to obtain
The corresponding predenoising processing result of data.
5. iterated denoising method according to claim 1, which is characterized in that the method packet of the adjustment objective degrees of confidence value
It includes: according to the difference of the confidence value for having acoustic frame unlabelled in the targeted voice signal and the current objective degrees of confidence value
Value, the current objective degrees of confidence value is tuned up or is turned down.
6. a kind of chip, which is characterized in that the chip is for storing any one of claim 1 to 5 iterated denoising method pair
The program code answered.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811512492.7A CN109346099B (en) | 2018-12-11 | 2018-12-11 | Iterative denoising method and chip based on voice recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811512492.7A CN109346099B (en) | 2018-12-11 | 2018-12-11 | Iterative denoising method and chip based on voice recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109346099A true CN109346099A (en) | 2019-02-15 |
| CN109346099B CN109346099B (en) | 2022-02-08 |
Family
ID=65303391
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811512492.7A Active CN109346099B (en) | 2018-12-11 | 2018-12-11 | Iterative denoising method and chip based on voice recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109346099B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458782A (en) * | 2019-08-15 | 2019-11-15 | 深圳市瑞立视多媒体科技有限公司 | Three-dimensional track smoothing processing method, device, equipment and storage medium |
| CN111935432A (en) * | 2020-08-12 | 2020-11-13 | 盛素杰 | Novel financial affairs are record for consultation device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002236497A (en) * | 2001-02-08 | 2002-08-23 | Alpine Electronics Inc | Noise reduction system |
| CN1905006A (en) * | 2005-07-27 | 2007-01-31 | 日本电气株式会社 | Noise suppression system, method and program |
| CN103177728A (en) * | 2011-12-21 | 2013-06-26 | 中国移动通信集团广西有限公司 | Method and device for conducting noise reduction on speech signals |
| CN103971680A (en) * | 2013-01-24 | 2014-08-06 | 华为终端有限公司 | Method and device for recognizing voices |
| CN108257617A (en) * | 2018-01-11 | 2018-07-06 | 会听声学科技（北京）有限公司 | A kind of noise scenarios identifying system and method |
| CN109410928A (en) * | 2018-12-11 | 2019-03-01 | 珠海市微半导体有限公司 | A kind of denoising method and chip based on speech recognition |
-
2018
- 2018-12-11 CN CN201811512492.7A patent/CN109346099B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002236497A (en) * | 2001-02-08 | 2002-08-23 | Alpine Electronics Inc | Noise reduction system |
| CN1905006A (en) * | 2005-07-27 | 2007-01-31 | 日本电气株式会社 | Noise suppression system, method and program |
| CN103177728A (en) * | 2011-12-21 | 2013-06-26 | 中国移动通信集团广西有限公司 | Method and device for conducting noise reduction on speech signals |
| CN103971680A (en) * | 2013-01-24 | 2014-08-06 | 华为终端有限公司 | Method and device for recognizing voices |
| CN108257617A (en) * | 2018-01-11 | 2018-07-06 | 会听声学科技（北京）有限公司 | A kind of noise scenarios identifying system and method |
| CN109410928A (en) * | 2018-12-11 | 2019-03-01 | 珠海市微半导体有限公司 | A kind of denoising method and chip based on speech recognition |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458782A (en) * | 2019-08-15 | 2019-11-15 | 深圳市瑞立视多媒体科技有限公司 | Three-dimensional track smoothing processing method, device, equipment and storage medium |
| CN110458782B (en) * | 2019-08-15 | 2022-05-03 | 深圳市瑞立视多媒体科技有限公司 | Three-dimensional track smoothing method, device, equipment and storage medium |
| CN111935432A (en) * | 2020-08-12 | 2020-11-13 | 盛素杰 | Novel financial affairs are record for consultation device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109346099B (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102687196B (en) | Method for the detection of speech segments | |
| US6535851B1 (en) | Segmentation approach for speech recognition systems | |
| JP6153142B2 (en) | Method for processing an acoustic signal | |
| CN111833902B (en) | Awakening model training method, awakening word recognition device and electronic equipment | |
| US20020143540A1 (en) | Voice recognition system using implicit speaker adaptation | |
| KR102333376B1 (en) | Speech recognition processing method for noise generating working device and system thereof | |
| CN111477219A (en) | Keyword distinguishing method and device, electronic equipment and readable storage medium | |
| CN117727298B (en) | Deep learning-based portable computer voice recognition method and system | |
| CN109545238A (en) | A kind of speech de-noising device and robot based on clean robot | |
| CN109346099A (en) | A kind of iterated denoising method and chip based on speech recognition | |
| CN111081223A (en) | Voice recognition method, device, equipment and storage medium | |
| JPWO2010128560A1 (en) | Speech recognition apparatus, speech recognition method, and speech recognition program | |
| CN112509557B (en) | Speech recognition method and system based on non-deterministic word graph generation | |
| CN109065026B (en) | Recording control method and device | |
| CN111105798B (en) | Equipment control method based on voice recognition | |
| CN109602333B (en) | Voice denoising method and chip based on cleaning robot | |
| CN109410928B (en) | Denoising method and chip based on voice recognition | |
| CN113539266A (en) | Command word recognition method and device, electronic equipment and storage medium | |
| US20240079004A1 (en) | System and method for receiving a voice command | |
| JP2002123286A (en) | Voice recognizing method | |
| CN109584899B (en) | Denoising device and cleaning robot based on voice recognition | |
| CN109360580B (en) | Iteration denoising device and cleaning robot based on voice recognition | |
| CN114822531A (en) | Liquid crystal television based on AI voice intelligent control | |
| CN115547352A (en) | Electronic device, method, apparatus and medium for processing noise thereof | |
| CN103390404A (en) | Information processing apparatus, information processing method and information processing program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 519000 2706, No. 3000, Huandao East Road, Hengqin new area, Zhuhai, Guangdong Applicant after: Zhuhai Yiwei Semiconductor Co.,Ltd. Address before: Room 105-514, No.6 Baohua Road, Hengqin New District, Zhuhai City, Guangdong Province Applicant before: AMICRO SEMICONDUCTOR Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |