CN109346075A - Identify user speech with the method and system of controlling electronic devices by human body vibration - Google Patents

Identify user speech with the method and system of controlling electronic devices by human body vibration Download PDF

Info

Publication number
CN109346075A
CN109346075A CN201811199154.2A CN201811199154A CN109346075A CN 109346075 A CN109346075 A CN 109346075A CN 201811199154 A CN201811199154 A CN 201811199154A CN 109346075 A CN109346075 A CN 109346075A
Authority
CN
China
Prior art keywords
human
vibration
sensor
signal
vibration sensor
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.)
Withdrawn
Application number
CN201811199154.2A
Other languages
Chinese (zh)
Inventor
林金锋
仇存收
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to CN201811199154.2A priority Critical patent/CN109346075A/en
Publication of CN109346075A publication Critical patent/CN109346075A/en
Priority claimed from PCT/CN2019/090883 external-priority patent/WO2019238061A1/en
Withdrawn legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00Speech recognition
    • G10L15/22Procedures used during a speech recognition process, e.g. man-machine dialogue
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00Speech recognition
    • G10L15/22Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G10L2015/223Execution procedure of a spoken command

Abstract

The embodiment of the present invention provide it is a kind of by human body vibration identify user speech with the system of controlling electronic devices, comprising: human body vibration sensor, for incuding the human body vibration of user;Processing circuit is coupled with the human body vibration sensor, for when the output signal for determining the human body vibration sensor includes user voice signal, control pick up facility to start pickup;Communication module is coupled with processing circuit and the pick up facility, for the communication between the processing circuit and the pick up facility.Another embodiment of the present invention provide it is a kind of by human body vibration identify user speech in the method for controlling electronic devices, comprising: detection human body vibration;When determining that the human body vibration includes vibration caused by user speaks, control pick up facility starts pickup.

Description

Identify user speech with the method and system of controlling electronic devices by human body vibration
Technical field
It is especially a kind of that the method for user speech is identified by human body vibration and is set the present invention relates to field of communication technology It is standby.
Background technique
With the progress of artificial intelligence (Artificial Intelligence, AI) technology, voice control technology is in hand On the consumption electronic products such as machine, tablet computer using more and more extensive.Now on the market there are many voice assistant product, example Such as the Siri of Apple Inc., the Google Assistant of Google, small ice of the Microsoft of Microsoft etc..These voice assistants Product is mounted in the terminal devices such as mobile phone, tablet computer, or is mounted in the intellectual products such as intelligent sound box, robot, is led to The phonetic order for crossing identification user executes corresponding operation, greatly convenient for users to use.
However, existing voice assistant there are problems that phonetic order source cannot be distinguished well in use.For example, Nearby if there is another person says phonetic order, it is possible to lead to the false triggering of voice assistant.In addition, in order to monitor The voice of user, to accomplish to respond at any time, the equipment for being mounted with voice assistant has to for a long time open microphone, causes function The rising of consumption.
Summary of the invention
The embodiment of the present invention provide it is a kind of by human body vibration identify user speech in the method for controlling electronic devices and System to reduce the false triggering of phonetic order, and reduces power consumption.
According to the first aspect of the invention, provide it is a kind of by human body vibration identify user speech with controlling electronic devices System, comprising:
Human body vibration sensor, for incuding the human body vibration of user;
Processing circuit is coupled with the human body vibration sensor, determines the defeated of the human body vibration sensor for working as When signal includes user voice signal out, control pick up facility starts pickup;
Communication module is coupled with processing circuit and the pick up facility, sets for the processing circuit and the pickup Communication between standby.
Optionally, the system also includes amplifier, the input terminal of the amplifier and the human body vibration sensors Output end is coupled, for amplifying the output signal of the human body vibration sensor.
Optionally, the system comprises high-pass filter, the input terminal of the high-pass filter and the human body vibrations to pass The output end of sensor is coupled, the low-frequency component of the output signal for filtering out the human body vibration sensor.
Optionally, the system comprises history buffers, the history buffers and the human body vibration sensor phase couplings It closes, the output signal for the human body vibration sensor in certain time before current time.
Optionally, the processing circuit includes voice analyzer, and the voice analyzer is for determining the human body vibration Whether the output signal of sensor includes user voice signal.
Optionally, the voice analyzer determines institute by analyzing the envelope of the output signal of the human body vibration sensor Whether the output signal for stating human body vibration sensor includes user voice signal;Alternatively,
The voice analyzer determines the human body by analyzing the frequency spectrum of the output signal of the human body vibration sensor Whether the output signal of vibrating sensor includes user voice signal.
Optionally, the system comprises ambient noise detectors, and the output for detecting the human body vibration sensor is believed Ambient noise in number.
Optionally, the system comprises false-alarm filter, the input terminal and the human body vibration of the false-alarm filter are passed Sensor is coupled, and output end is coupled with the voice analyzer, for filtering the output signal of the human body vibration sensor In the noise from inside of human body.
Optionally, the processing circuit further includes training module, and the training module is coupled with the voice analyzer, Speech analysis model for training the voice analyzer to use.
Optionally, the voice analyzer includes the software module operated on the processing circuit or processing electricity Hardware module in road.
The optional training module includes in the software module gone on the processing circuit or the processing circuit Hardware module.
Optionally, the system also includes memory, the memory and the training module and the voice analyzers It is coupled, the speech analysis model generated for storing the training module.
Optionally, the human body vibration sensor includes bone vibrating sensor.
According to another aspect of the present invention, provide it is a kind of by human body vibration identify user speech with controlling electronic devices Method, comprising:
Detect human body vibration;
When determining that the human body vibration includes vibration caused by user speaks, control pick up facility starts pickup.
Optionally, the pick up facility is usually to close, the method also includes:
Determining the human body vibration does not include when vibrating caused by user speaks, the pick up facility being maintained to close.
Optionally, after the detection human body vibration, the determination human body vibration includes caused by user speaks When vibration, before control pick up facility starts pickup, the method also includes:
Filter out the noise in the human body vibration.
Optionally, the human body vibration includes bone vibration.
The system and method provided according to embodiments of the present invention, speech recognition system only can be opened just when user speaks Pickup reduces the situation that speech recognition system misidentifies other people phonetic orders in this way, and further, pick up facility can only exist Unlatching when user speaks reduces the power consumption of system.
Detailed description of the invention
Figure 1A is the specific form figure of one embodiment of the invention;
Figure 1B is the structural schematic diagram of one embodiment of the invention;
Fig. 2A is the ken figure and spectrogram for the voice signal that microphone picks up;
Fig. 2 B is the ken figure and spectrogram for the voice signal that bone conduction sensor picks up;
Fig. 3 is the structural schematic diagram of another embodiment of the present invention;
Fig. 4 is the signal that detects of bone conduction sensor by the comparison diagram before and after noise reduction process;
Fig. 5 is a spectrogram of human speech;
Fig. 6 is the time-domain diagram that the mankind chew noise;
Fig. 7 is the flow chart of an embodiment of the method for the invention.
Specific embodiment
The embodiment of the present invention is described in detail below in conjunction with attached drawing.
As shown in Figure 1B, one embodiment of the present of invention provides a kind of system for identifying user speech, which can position In in the same equipment, such as earphone, hearing aid or other dedicated equipment;It can also be distributed on different devices, such as A part is on earphone, and another part, on the electronic equipments such as mobile phone, intelligent sound box, the embodiment of the present invention does not limit this It is fixed.The equipment includes:
Human body vibration sensor 110 (bone sensor), for incuding the human body vibration of user;
Processing circuit 120 is coupled with the human body vibration sensor 110, for when the determining human body vibration sensing When the output signal of device includes user voice signal, control pick up facility starts pickup;
Communication module 130 is coupled with processing circuit 120 and pick up facility 140, sets for processing circuit 120 and pickup Communication between standby 140;
Pick up facility 140, for picking up sound.
As shown in Figure 1A, the system of the identification user speech can be with electronic equipment 200 (such as mobile phone, tablet computer, intelligence Speaker, robot) it is used cooperatively, speech recognition system is installed, the speech recognition system can be on the electronic equipment 200 It identifies the voice of user, and can be further processed, such as executed as electronic equipment 200 described in voice command control Corresponding operation.
Human body vibration sensor 110 is the sensor for incuding human body vibration.People can be served as there are many kinds of sensor Body vibrating sensor 110, such as bone conduction sensor.Bone conduction sensor is a kind of vibration for incuding bone, and by the vibration Be converted to the device of electric signal, optical signal or other signals.As a kind of medium, bone can propagate sound as air Wave can cause the vibration of bone when sound wave is propagated in bone.There are many kinds of bone conduction sensors, and the embodiment of the present invention can be selected With the 13x2 type sensor of existing bone conduction sensor, such as Sonion company, it is preferred that can choose sampling bandwidth 2kHz Or above bone conduction sensor, sensitivity can achieve -34dB.Bone conduction sensor may be mounted at close to skeleton Position may be mounted in the earphone heads of earphone as shown in Figure 1A, and when user wears earphone, earphone heads protrude into ear canal, sensor It can detect the bone vibration transmitted from ear canal.Certainly, sensor also may be mounted at other positions, the embodiment of the present invention pair This is not construed as limiting.In the implementation of the present invention, inventors have found that since the spread speed of sound in bone will be faster than sky Spread speed in gas, therefore bone conduction sensor can sense that user speaks the vibration of generation earlier, so as to and When judge that user is speaking, as far as possible reduction microphone open time delay.
Simultaneously in addition to very strong ambient noise, general environment noise is difficult to generate strong vibration in bone, therefore Compared with general microphone, the signal that bone vibrating sensor detects is purer human voice signal, such as Fig. 2A and Fig. 2 B institute Show, Fig. 2A is the voice signal of a common microphone record, it can be seen that a large amount of ambient noise is wherein contained, such as The ingredient of many high frequencies, and Fig. 2 B indicates the voice signal of bone conduction sensor record, it can be seen that compared with microphone, The signal of bone conduction sensor record will more " pure ", and the noise signal of various high frequencies obviously disappears.Therefore, bone vibrating sensing The more difficult interference by ambient noise of device.
Certainly, in the embodiment of the present invention, human body vibration sensor is also possible to other sensors, such as is attached to human body Acceleration transducer on skin can perceive the vibration of skin;Or it is connected to the biopotential sensor on human body, such as various Electrode can perceive the biological Electrical change of human body to detect biology Electrical change caused by human body vibration.Implementation of the invention Example is not construed as limiting this.
Processing circuit 120 can be central processing unit (central to be any circuit with processing function Processing unit, CPU), digital signal processor (digital signal processor, DSP) or dedicated place Manage device.In one embodiment, processor DSP, such as the DA14195 type DSP of Dialog Semiconductor company.Place Reason circuit 120 can integrate on one chip, can also be dispersed on several pieces of chips, can also be completely separate circuit Element is composed.Processing circuit 120 is responsible for the signal that processing bone conduction sensor detects, and controls whole system.
Communication module 130 is in processing circuit 120 and system and/or the device outside system, particularly pick up facility 140 It is communicated.For example, when pick up facility 140 is located in another equipment (for example processor 120 is located on earphone, and microphone 140 are located on mobile phone), it can be communicated between processor and microphone by communication module 130.Communication module 130 can To be wire communication module or wireless communication module, for example, Wireless Fidelity (WiFi) module, bluetooth (Bluetooth) module, Near-field communication (near field communication, NFC) module etc., the embodiment of the present invention is not construed as limiting this.
In some embodiments, above-mentioned communication module 130 is also possible to simple conducting wire, is used for transmission processing circuit 120 With the signal between the device in system and/or outside system.
Pick up facility 140 is equipment for picking up sound, such as microphone, microphone etc..
In some embodiments, pick up facility may not be a part of the system provided by the invention.
In some embodiments of the invention, the human body vibration signal that human body vibration sensor will test is transmitted to processing electricity Road controls the pick up facility pickup when processing circuit determines that the human body vibration signal is vibration caused by user speaks.It is logical This process is crossed, speech recognition system only can just open pickup when user speaks, and reduce speech recognition system mistake in this way Identify the situation of other people phonetic orders, in some embodiments, the pick up facility only unlatching when user speaks reduces system Power consumption.
How system provided in an embodiment of the present invention described further below realizes the differentiation to user speech.
As shown in figure 3, in some embodiments of the invention, human body vibration sensor 110 is specially bone conduction sensor 110, bone conduction sensor 110 is coupled with amplifier 310, and amplifier 310 is coupled with analog-digital converter 320, analog-to-digital conversion Device 320 is coupled with high-pass filter (high-pass filter) 330, high-pass filter 330 and history buffers (history buffer) 340 is coupled, the output end and ambient noise detector (ambient noise of history buffers 340 Detector it) 350 is coupled, ambient noise detector 350 is coupled with voice analyzer (envelope detector) 390. In some embodiments, the output end of history buffers and automatic gain controller 370 are coupled, automatic gain controller 370 Then it is coupled with false-alarm filter (false alert filter) 380, and false-alarm filter 380 is then coupled to voice analyzer 390。
Illustrate processing of the system to signal below with reference to signal flow.The signal that bone conduction sensor 110 generates is through over-discharge The amplification of big device 310, is transferred to analog-digital converter 320, and the simulation that analog-digital converter 320 generates bone conduction sensor 110 is believed Number it is converted into digital signal, is transferred to high-pass filter 330.The function of high-pass filter 330 is to filter out direct current signal and low frequency Noise, by the filtering of high-pass filter 330, high-frequency signal is extracted, into history buffers 340.History buffers 340 function is the signal before caching current time in some time, the so subsequent letter for needing to handle in this period It number can.By this caching, actually play the role of to above-mentioned signal framing, that is to say, that signal is split For several segments, handled as unit of these segments.In some embodiments, 2 milliseconds before history buffers 340 cache Interior signal, inventor have found that this value can preferably guarantee the timely starting of microphone in the implementation of the present invention, Such as microphone can be started in 50 milliseconds after user loquiturs.
In some embodiments, the signal that history buffers 340 export enters ambient noise detector 350.Ambient noise Detector 350 is used to detect the ambient noise in above-mentioned signal, and filters out the ambient noise.Certainly, ambient noise detector 350 The ambient noise can not also be filtered out, and the information of ambient noise signal is only transferred to subsequent processing.Here, ambient noise Refer to the noise in surrounding enviroment, such as user's sound etc. that other people one's voices in speech, other objects issue at one's side.Usually come It says, the vibration of osteoacusis is mostly caused by inside of human body vibration source, such as speaks, heartbeat, chewing, walks etc., external environment Noise is not easy to cause detectable vibration in skeleton in the lower situation of volume, but works as external environment very noise When miscellaneous (such as outside noise reaches 80dB), the noise of external environment may also cause detectable vibration in bone, thus Voice signal is interfered.
The noise reduction process method of ambient noise detector 350 has very much, has in traditional technology much to microphone denoising Technology, such as dual microphone noise reduction, in the art, two microphones for being arranged at a distance pick up sound respectively, One of them sound source close to voice, another sound source far from voice, the former voice signal that picks up is more in this way, and the latter The noise of pickup is more, and two compare, so that it may filter out noise.In some embodiments of the invention, it can be passed far from bone Microphones for picking up noises, or the bone conduction sensor for picking up noise are arranged in the position of derivative sensor 110, in this way The technology filtering environmental noise of dual microphone noise reduction can be used.
The implementation of another lower cost is then the characteristic by analyzing ambient noise, filtering environmental noise.Example Such as, frequency spectrum, intensity, duration, stationary noise or the nonstationary noise of ambient noise are detected.Typically, people speaks Fundamental frequency is about in the range of 500-1000Hz, and the frequency multiplication upper limit is about in 3000Hz, higher frequency multiplication for speech processes and Speech be it is unnecessary, noise can be considered as.And on the duration, ambient noise is usually lasting existing, that is, if continued Bone conduction sensor is opened, then can constantly receive ambient noise signal, and in contrast, user speaks the language of generation Sound signal is then intermittent, because user will not always speak.Optionally, mould can be carried out with the model of use environment noise Formula identification, isolates noise from signal.It can also be by the way that all signals except voice signal be all considered as ambient noise Mode isolates noise signal.In some embodiments, the system also includes threshold selectors 355, with ambient noise Detector 350 and training module 360 are coupled, for select determine noise " thresholding ", such as determine a signal be noise Model parameter critical value.
In one particular embodiment of the present invention, after the noise reduction process that have passed through ambient noise detector 350, environment is made an uproar The signal of sound greatly weakens, as shown in figure 4, the signal time-domain diagram in the figure before the first behavior noise reduction, before third behavior noise reduction Signal spectrum figure, the signal time-domain diagram after the second behavior noise reduction, fourth line are the signal spectrum figure after noise reduction, in two rectangle frames Signal be the lower signal of period record that user speaks, it can be seen that by noise reduction process, speak twice between noise signal It is removed, and the noise signal during speaking also greatly weakens, voice signal is opposite to be enhanced.
In some embodiments, voice analyzer 390 is used to analyze the envelope of above-mentioned signal, so that picking out the signal is The no voice for user.As it is known by the man skilled in the art, the waveform envelope and noise of human speech are different, usually come It says, the speech waveform envelope that different people is spoken also is different.Voice analyzer 390 can identify the letter according to the envelope of signal It number is voice or noise.
In further embodiments, voice analyzer 390 can analyze the frequency spectrum of above-mentioned signal, to pick out the signal Whether be user voice.As it is known by the man skilled in the art, the frequency spectrum of human speech and the frequency spectrum of noise are different, such as scheme Shown in 5, the vowel of human language and voiced consonant have specific formant on frequency domain, by the mould for identifying these formants Formula, so that it may differentiate that above-mentioned signal belongs to human speech or noise.
The method whether identification signal belongs to voice signal or identify voice owner has very much, in some embodiments In, it can be used user's training pattern to carry out pattern match to signal, determine whether signal belongs to user speech.Pattern match The technology of various field of speech recognition can be used, such as Hidden Markov Model etc., this will not be repeated here.
As previously mentioned, when the signal that voice analyzer 390 determines that bone vibrating sensor transmits includes user speech, it is described Processing circuit 120, which generates, controls the control signal that the pick up facility 140 starts pickup.
In certain embodiments, the signal that history buffers 340 export enters false-alarm filter 380, false-alarm filter 380 For filtering the noise from inside of human body, such as by the noise for the generations such as chew, walk.The signal enters speech analysis later Device 390.The processing of voice analyzer 390 is already described above, and details are not described herein again.
The filtering of inside of human body noise signal can be carried out based on the features such as frequency spectrum, intensity, period.Such as chewing Signal characteristic as shown in fig. 6, can establish accordingly chewing signal model, chewing noise is identified by pattern match.
The characteristics of chewing noise and filtering method, can refer to " Mastication noise reduction method For fully implantable hearing aid using piezo-electric sensor " (Sung Dae Na etc. People is published in 25 (2017) S29-S34 of Technology and Health Care) etc. documents.
As shown in figure 3, automatic gain controller (automatic gain can be set before false-alarm filter 380 Controller, AGC) function of 370, AGC 370 is adjusted to the intensity of signal, the signal normalization of varying strength To the intensity of a standard.The difference of the position of the volume, bone conduction sensor and the human contact that are spoken due to user etc. can be led It causes the signal strength of bone conduction sensor output different, in order to facilitate subsequent processing, its intensity can be normalized.
As a kind of more preferably embodiment, as shown in figure 3, training module 360, the module and ring can be arranged in systems Border noise detector 350 and/or voice analyzer 390 are coupled, and carry out for the model to ambient noise and/or user speech Training.By training, the accuracy of above-mentioned model can be promoted, to improve the judgement to ambient noise and/or user speech Precision.Optionally, training module can also be coupled with false-alarm filter 380, for training the model of inside of human body noise.It is right The method that machine learning can be used in the training of model, such as be trained by neural network.
Optionally, above system further includes memory 365, with 390 phase of the training module 360 and the voice analyzer Coupling, for storing the model of training generation.
It will be understood by those skilled in the art that the various pieces of above system were not necessarily required to.For example, if bone The output signal strength of conduction sensor 110 can satisfy the needs of follow-up signal processing, then amplifier 310 is not just required 's.If of less demanding to filtering out for ambient noise (as previously mentioned, in addition to very noisy environment, general bone conduction sensor It is difficult to sense extraneous environmental noise), then ambient noise detector 350 is nor required.
Amplifier, analog-digital converter, high-pass filter, history buffers, ambient noise detector, door in above system Limit selection, automatic gain controller, false-alarm filter, training module and voice analyzer can be respectively discrete device, It can be and be partly or completely integrated in one piece or several pieces of chips.The ambient noise detector, thresholding select, certainly Dynamic gain controller, false-alarm filter, training module and voice analyzer, can be hardware circuit respectively or operate in processing Software module on circuit 120.In a specific embodiment, above-mentioned analog-digital converter, high-pass filter, history buffer Device, ambient noise detector, thresholding selection, automatic gain controller, false-alarm filter, training module and voice analyzer all positions In in a DSP.Optionally, above-mentioned ambient noise detector, thresholding selection, automatic gain controller, false-alarm filter, training Module and voice analyzer are the software module operated on the DSP.
As shown in fig. 7, another embodiment of the present invention provides a kind of identification user speech with the side of controlling electronic devices Method, comprising:
710, human body vibration is detected;
Detect human body vibration method in front it is stated that, details are not described herein again.In certain embodiments, it is shaken by bone Dynamic sensor detection bone vibration.
750, when determining that the human body vibration includes vibration caused by user speaks, control pick up facility starts pickup.
In some embodiments, the pick up facility is usually to close, the method also includes:
760, determining the human body vibration does not include when vibrating caused by user speaks, the pick up facility being maintained to close.
Determine whether the human body vibration includes the specific method vibrated caused by user speaks, describes voice point in front It has been described in detail when parser 390, details are not described herein.
As previously mentioned, in certain embodiments, after described 710, before 750, the method also includes:
720, the noise in the human body vibration is filtered out;
Filter out the method for noise in front it is stated that, details are not described herein again.
In addition, being said respectively in each technology illustrated respectively in above embodiments, system, device, method and each embodiment Bright technical characteristic can be combined, so that the other modules not departed within the spirit and principles in the present invention are formed, side Method, device, system and technology, these are according to an embodiment of the present invention to record the module that is composed, method, device, system and Technology is within the scope of the present invention.
Obviously, those skilled in the art should be understood that above-mentioned each unit of the invention or each step can be with general Computing device realize that they can be concentrated on a single computing device, or be distributed in multiple computing devices and formed Network on, optionally, they can be realized with the program code that computing device can perform, it is thus possible to which they are stored It is performed by computing device in the storage device.Perhaps they are fabricated to each circuit modules or will be in them Multiple units or step are fabricated to single circuit module to realize.In this way, the present invention is not limited to any specific hardware and soft Part combines.
It is above presently preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all of the invention Any modification, equivalent replacement, improvement and so within spirit and principle, are included within the scope of protection of the present invention.

Claims (17)

1. a kind of identify user speech with the system of controlling electronic devices by human body vibration, comprising:
Human body vibration sensor, for incuding the human body vibration of user;
Processing circuit is coupled with the human body vibration sensor, for when the output letter for determining the human body vibration sensor Number include user voice signal when, control pick up facility start pickup;
Communication module is coupled with processing circuit and the pick up facility, for the processing circuit and the pick up facility it Between communication.
2. the system as claimed in claim 1, which is characterized in that further include amplifier, the input terminal of the amplifier with it is described The output end of human body vibration sensor is coupled, for amplifying the output signal of the human body vibration sensor.
3. system as claimed in claim 1 or 2, which is characterized in that including high-pass filter, the input of the high-pass filter End and the output end of the human body vibration sensor are coupled, for filter out the human body vibration sensor output signal it is low Frequency ingredient.
4. system as described in any one of claims 1-3, which is characterized in that including history buffers, the history buffers It is coupled with the human body vibration sensor, for the defeated of the human body vibration sensor in certain time before current time Signal out.
5. system according to any one of claims 1-4, which is characterized in that the processing circuit includes voice analyzer, institute Voice analyzer is stated for determining whether the output signal of the human body vibration sensor includes user voice signal.
6. system as claimed in claim 5, it is characterised in that:
The voice analyzer determines the human body vibration by analyzing the envelope of the output signal of the human body vibration sensor Whether the output signal of sensor includes user voice signal;Alternatively,
The voice analyzer determines the human body vibration by analyzing the frequency spectrum of the output signal of the human body vibration sensor Whether the output signal of sensor includes user voice signal.
7. such as system described in claim 5 or 6, which is characterized in that including ambient noise detector, for detecting the human body Ambient noise in the output signal of vibrating sensor.
8. such as the described in any item systems of claim 5-7, which is characterized in that including false-alarm filter, the false-alarm filter Input terminal be coupled with the human body vibration sensor, output end is coupled with the voice analyzer, described for filtering From the noise of inside of human body in the output signal of human body vibration sensor.
9. such as the described in any item systems of claim 5-8, which is characterized in that the processing circuit further includes training module, institute It states training module to be coupled with the voice analyzer, the speech analysis model for training the voice analyzer to use.
10. such as system described in claim 5 or 6, which is characterized in that the voice analyzer includes operating in the processing electricity Hardware module in the software module of road or the processing circuit.
11. system as claimed in claim 9, which is characterized in that the training module includes row on the processing circuit Hardware module in software module or the processing circuit.
12. such as the described in any item systems of claim 1-11, which is characterized in that it further include memory, the memory and institute It states training module and the voice analyzer is coupled, the speech analysis model generated for storing the training module.
13. such as the described in any item systems of claim 1-12, which is characterized in that the human body vibration sensor includes bone vibration Sensor.
14. a kind of identify user speech in the method for controlling electronic devices by human body vibration, comprising:
Detect human body vibration;
When determining that the human body vibration includes vibration caused by user speaks, control pick up facility starts pickup.
15. method as claimed in claim 14, which is characterized in that the pick up facility is usually to close, and the method is also Include:
Determining the human body vibration does not include when vibrating caused by user speaks, the pick up facility being maintained to close.
16. the method as described in claims 14 or 15, which is characterized in that after the detection human body vibration, the determination When the human body vibration includes vibration caused by user speaks, before control pick up facility starts pickup, the method also includes:
Filter out the noise in the human body vibration.
17. such as the described in any item methods of claim 14-16, which is characterized in that the human body vibration includes bone vibration.
CN201811199154.2A 2018-10-15 2018-10-15 Identify user speech with the method and system of controlling electronic devices by human body vibration Withdrawn CN109346075A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811199154.2A CN109346075A (en) 2018-10-15 2018-10-15 Identify user speech with the method and system of controlling electronic devices by human body vibration

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811199154.2A CN109346075A (en) 2018-10-15 2018-10-15 Identify user speech with the method and system of controlling electronic devices by human body vibration
PCT/CN2019/090883 WO2019238061A1 (en) 2018-06-12 2019-06-12 Method and device for recognizing user voice by means of human body vibration

Publications (1)

Publication Number Publication Date
CN109346075A true CN109346075A (en) 2019-02-15

Family

ID=65308771

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811199154.2A Withdrawn CN109346075A (en) 2018-10-15 2018-10-15 Identify user speech with the method and system of controlling electronic devices by human body vibration

Country Status (1)

Country Link
CN (1) CN109346075A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110191388A (en) * 2019-05-31 2019-08-30 深圳市荣盛智能装备有限公司 Bone conduction earphone noise-reduction method, device, electronic equipment and storage medium
CN110191387A (en) * 2019-05-31 2019-08-30 深圳市荣盛智能装备有限公司 Automatic starting control method, device, electronic equipment and the storage medium of earphone
CN110265007A (en) * 2019-05-11 2019-09-20 出门问问信息科技有限公司 Control method, control device and the bluetooth headset of voice assistant system
WO2019238061A1 (en) * 2018-06-12 2019-12-19 华为技术有限公司 Method and device for recognizing user voice by means of human body vibration
CN110931031A (en) * 2019-10-09 2020-03-27 大象声科(深圳)科技有限公司 Deep learning voice extraction and noise reduction method fusing bone vibration sensor and microphone signals
WO2021008458A1 (en) * 2019-07-12 2021-01-21 Guangdong Oppo Mobile Telecommunications Corp., Ltd. Method for voice recognition via earphone and earphone
WO2021068120A1 (en) 2019-10-09 2021-04-15 大象声科(深圳)科技有限公司 Deep learning speech extraction and noise reduction method fusing signals of bone vibration sensor and microphone

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006025079A (en) * 2004-07-07 2006-01-26 Nec Tokin Corp Head set and wireless communication system
CN101042869A (en) * 2006-03-24 2007-09-26 致胜科技股份有限公司 Nasal bone conduction living body sound-groove identification apparatus
CN104144377A (en) * 2013-05-09 2014-11-12 Dsp集团有限公司 Low power activation of voice activated device
CN105453174A (en) * 2013-06-03 2016-03-30 三星电子株式会社 Speech enhancement method and apparatus for same

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006025079A (en) * 2004-07-07 2006-01-26 Nec Tokin Corp Head set and wireless communication system
CN101042869A (en) * 2006-03-24 2007-09-26 致胜科技股份有限公司 Nasal bone conduction living body sound-groove identification apparatus
CN104144377A (en) * 2013-05-09 2014-11-12 Dsp集团有限公司 Low power activation of voice activated device
CN105453174A (en) * 2013-06-03 2016-03-30 三星电子株式会社 Speech enhancement method and apparatus for same

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019238061A1 (en) * 2018-06-12 2019-12-19 华为技术有限公司 Method and device for recognizing user voice by means of human body vibration
CN110265007A (en) * 2019-05-11 2019-09-20 出门问问信息科技有限公司 Control method, control device and the bluetooth headset of voice assistant system
CN110265007B (en) * 2019-05-11 2020-07-24 出门问问信息科技有限公司 Control method and control device of voice assistant system and Bluetooth headset
CN110191388A (en) * 2019-05-31 2019-08-30 深圳市荣盛智能装备有限公司 Bone conduction earphone noise-reduction method, device, electronic equipment and storage medium
CN110191387A (en) * 2019-05-31 2019-08-30 深圳市荣盛智能装备有限公司 Automatic starting control method, device, electronic equipment and the storage medium of earphone
WO2021008458A1 (en) * 2019-07-12 2021-01-21 Guangdong Oppo Mobile Telecommunications Corp., Ltd. Method for voice recognition via earphone and earphone
CN110931031A (en) * 2019-10-09 2020-03-27 大象声科(深圳)科技有限公司 Deep learning voice extraction and noise reduction method fusing bone vibration sensor and microphone signals
WO2021068120A1 (en) 2019-10-09 2021-04-15 大象声科(深圳)科技有限公司 Deep learning speech extraction and noise reduction method fusing signals of bone vibration sensor and microphone

Similar Documents

Publication Publication Date Title
CN109346075A (en) Identify user speech with the method and system of controlling electronic devices by human body vibration
CN104144377B9 (en) The low-power of voice activation equipment activates
CN105765656B (en) Control the speech recognition process of computing device
CN106782591B (en) Device and method for improving speech recognition rate under background noise
CN107481718B (en) Audio recognition method, device, storage medium and electronic equipment
JP2004199053A (en) Method for processing speech signal by using absolute loudness
CN102697520B (en) Electronic stethoscope based on intelligent distinguishing function
CN110268470A (en) The modification of audio frequency apparatus filter
CN106664473A (en) Information-processing device, information processing method, and program
CN104216677A (en) Low-power voice gate for device wake-up
CN101023469A (en) Digital filtering method, digital filtering equipment
CN108681440A (en) A kind of smart machine method for controlling volume and system
US10347249B2 (en) Energy-efficient, accelerometer-based hotword detection to launch a voice-control system
JP2013142843A (en) Operation analyzer, voice acquisition device, and operation analysis system
CN108461081A (en) Method, apparatus, equipment and the storage medium of voice control
GB2526980A (en) Sensor input recognition
CN109036395A (en) Personalized speaker control method, system, intelligent sound box and storage medium
JP2016535305A (en) A device for improving language processing in autism
CN110728993A (en) Voice change identification method and electronic equipment
GB2553040A (en) Sensor input recognition
WO2019238061A1 (en) Method and device for recognizing user voice by means of human body vibration
JP6051996B2 (en) Speech analysis apparatus, speech analysis system and program
CN109920419B (en) Voice control method and device, electronic equipment and computer readable medium
CN208538474U (en) Speech recognition system
TWI730584B (en) Keyword detecting method and associated device

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
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20190215