CN105976831A - Lost child detection method based on cry recognition - Google Patents
Lost child detection method based on cry recognition Download PDFInfo
- Publication number
- CN105976831A CN105976831A CN201610314228.7A CN201610314228A CN105976831A CN 105976831 A CN105976831 A CN 105976831A CN 201610314228 A CN201610314228 A CN 201610314228A CN 105976831 A CN105976831 A CN 105976831A
- Authority
- CN
- China
- Prior art keywords
- child
- sound
- frame
- volume
- detection
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 35
- 238000012706 support-vector machine Methods 0.000 claims abstract description 14
- 238000000034 method Methods 0.000 claims abstract description 9
- 238000001228 spectrum Methods 0.000 claims description 16
- 230000005236 sound signal Effects 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 claims description 2
- 238000001914 filtration Methods 0.000 claims 1
- 238000007689 inspection Methods 0.000 claims 1
- 206010011469 Crying Diseases 0.000 abstract description 4
- 238000012544 monitoring process Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/21—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being power information
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Emergency Alarm Devices (AREA)
- Alarm Systems (AREA)
Abstract
The invention relates to a lost child detection method based on cry recognition, and the method achieves the reliable detection of a lost child in a double-layer detection mode of volume initial detection and sound content rechecking. The method comprises the steps: firstly detecting whether there is large-volume sound in a school bus or not through employing short-time energy features: determining that there may be the lost child if there is the large-volume sound in the school bus; secondly combining the Mel-frequency cepstral coefficient features and a support vector machine classifier to carry out sound content recheck, determining whether there is a child crying and shouting or not, and reliably detecting whether there is the lost child in the school bus or not.
Description
Technical field
The present invention relates to a kind of omission child's detection method, belong to safety monitoring technical field.
Background technology
Omit child on school bus abnormally dangerous, easily cause child's death by accident.In order to improve the safety of school bus operation,
Ensure child's safety, except to responsible person concerned with in addition to child carries out regular safety education and supervision and check by bus, and also
Need to improve discovery and the pre-alerting ability omitting child technically.Patent " is detained occupant state to identify and danger state control in car
System " (numbering: CN201510779628.0,2015) by whether there are passenger on pressure transducer detection seat leave over,
Detection passenger is left over certain effectiveness, but child has been omitted for detection and acts on not quite, because this detection method cannot be distinguished by
People and thing, also cannot detect corridor or other do not install the child that leaves over of sensor region, and detection leakage phenomenon is serious.Examine based on sound
The research surveying child's sob is in progress substantially at present, such as document " design of sob detection module and realization in baby monitoring system " (number
Word technology and application, 2014) devise the detection device of a kind of vagitus, can according to the frequency detecting sob of sound, but
Unsatisfactory for detection omission child's effect, one is because being also possible to point in addition to crying and shout when omission child finds danger
Crying, two is that the sensitivity of this device is the highest.
Summary of the invention
Spy of the present invention proposes a kind of omission child's detection method reliably.
What the technical problem to be solved was to provide on a kind of school bus leaves over child's detection method, initially with
Whether there is, in short-time energy feature detection school bus, the sound given great volume, if it is present think and there may be omission child；So
Rear combination mel cepstrum coefficients feature and support vector machine classifier carry out sound-content and check, and are confirmed whether to there is crying of child
Yell and shriek, whether reliable detection school bus exists omission child.
Suspicious big volume whether is there is, it is to avoid leakage in using the short-time energy feature Preliminary detection school bus of acoustical signal
Alert；Using mel cepstrum coefficients and support vector machine to check sound-content, reduce false-alarm, final raising is believed according to audio frequency
Number detection omit child reliability, can be widely used for safety of school bus field.
For achieving the above object, the present invention uses following technical scheme, flow chart as shown in Figure 1:
1, volume initial survey based on short-time energy
After school bus stops, owing to vehicle window is closed, in car, sound is the least.If there is omitting child, child finds danger
Sob and shriek after danger are the biggest.Based on this thinking, the present invention is first depending on the volume of pick up collection and comes
Whether Preliminary detection exists omission child.Needing exist for explanation, voice data presses frame input, sample frequency in the present invention
For 8KHz, every 256 sampled points are a frame, the most overlapping between frame with frame.Volume initial survey step is:
Step1: fast Fourier transform
IfRepresent thetFramenThe audio signal of individual sampled point, this signal is the linear combination of pure sound and noise,
In order to filter noise, need to be smoothed at frequency domain.Therefore it is the most rightCarry out fast Fourier transform, obtain amplitude spectrum, k represents frequency indices.
Step2: smooth power spectrum calculates
The smooth power spectrum of t frame acoustical signal can be expressed as
Wherein,For smoothing factor, in the present invention, take=0.5。Represent the smooth power spectrum of t-1 frame, initially
Smooth power spectrumIt is set to 0.
Step3: Fourier inversion
RightCarry out Fourier inversion, obtain filtered audio signal。
Step4: short-time energy calculates
The short-time energy of t frameCan be expressed as
Step5: volume initial survey
IfMore than threshold value T, then it is assumed that there may be omission child, enter sound-content and check the stage.Otherwise, continue into
Row volume initial survey.In view of system just start a bit of time Nei Chenei of (namely car door just close) be noiseless (because this
Time school bus on people do not walk far, if sound, can cause other people attention, rescue omits child), in the present invention,
After taking system start-up, the meansigma methods of front 50 frame short-time energies is as the value of threshold value T.
2, sound-content based on mel cepstrum coefficients and support vector machine is checked
Simple dependence volume detection is omitted child and is generally not present false dismissal, but false alarm rate is the highest.For reducing false alarm rate, the present invention exists
After volume initial survey finds suspicious object, then use sound identification method to analyze sound-content, distinguish whether sound exists and cry
Yell, shriek, this two classes sound is to omit the emergency mode that child finds generally to use after danger.Concretely comprise the following steps:
Step1: ask for logarithmic energy spectrum
Amplitude spectrum to t frame audio signalTake the logarithm, obtain logarithmic energy spectrum。
Step2: calculate mel cepstrum coefficients
The computing formula of mel cepstrum coefficients (MFCC) is:
Wherein,Representing the n-th mel cepstrum coefficients of t frame, K represents the number of Mel (Mel) bank of filters.At this
In invention, K takes 24.Giving up flip-flop, the present invention takes n=1, and 2 ..., 12.
Step3:SVM classifies
Using 12 mel cepstrum coefficients as feature, input support vector machine (SVM) grader is classified, if classification results
It is 1, then it is assumed that there is omission child, now start alarm device, alert information is sent to school bus director and higher level's supervision
Department, reminds related personnel to rescue in time.SVM classifier needs precondition to obtain, and the training method of the present invention is: select
100 sections comprise child cry and shout, the sound clip screamed as positive sample, and 1000 sections of other sound clips are as negative sample,
Extract the mel cepstrum coefficients feature of Different categories of samples, select RBF as kernel function, the SVM instrument provided with Matlab
Bag is trained, and obtains SVM classifier.
It is an advantage of the current invention that: use the double-deck detection pattern that volume initial survey and sound-content are checked, it is achieved omit child
Virgin reliable detection.
Accompanying drawing explanation
Fig. 1 leaves over child's overhaul flow chart.
Detailed description of the invention
Leave over child's detection method on a kind of school bus, whether there is sound initially with in short-time energy feature detection school bus
Measure big sound, if it is present think and there may be omission child；Then in conjunction with mel cepstrum coefficients feature and support vector
Whether machine grader carries out sound-content and checks, and is confirmed whether to there is the cry of child and shriek, on reliable detection school bus
Exist and omit child.
Suspicious big volume whether is there is, it is to avoid leakage in using the short-time energy feature Preliminary detection school bus of acoustical signal
Alert；Using mel cepstrum coefficients and support vector machine to check sound-content, reduce false-alarm, final raising is believed according to audio frequency
Number detection omit child reliability, can be widely used for safety of school bus field.
Claims (2)
1. based on sob identification leave over child's detection method, it is characterised in that initially with short-time energy feature detection school bus
The most whether there is the sound given great volume, if it is present think and there may be omission child；Special then in conjunction with mel cepstrum coefficients
Support vector machine classifier of seeking peace carries out sound-content and checks, and is confirmed whether to exist the cry of child and shriek, detects school
Omission child whether is there is on car；
Flow process is as follows:
(1), volume initial survey based on short-time energy
The volume being first depending on pick up collection comes whether Preliminary detection exists omission child, in the present invention voice data
Inputting by frame, sample frequency is 8KHz, and every 256 sampled points are a frame, the most overlapping between frame with frame, and volume initial survey step is:
Step1: fast Fourier transform
IfRepresent thetFramenThe audio signal of individual sampled point, this signal is the linear combination of pure sound and noise,
For filtering noise, need to be smoothed at frequency domain, the most rightCarry out fast Fourier transform, obtain amplitude spectrum, k
Represent frequency indices；
Step2: smooth power spectrum calculates
The smooth power spectrum of t frame acoustical signal is expressed as
Wherein,For smoothing factor, in the present invention, take=0.5,Represent the smooth power spectrum of t-1 frame, initially put down
Sliding power spectrumIt is set to 0；
Step3: Fourier inversion
RightCarry out Fourier inversion, obtain filtered audio signal；
Step4: short-time energy calculates
The short-time energy of t frameCan be expressed as
Step5: volume initial survey
IfMore than threshold value T, then it is assumed that there may be omission child, enter sound-content and check the stage, otherwise, proceed
Volume initial survey；In the present invention, after taking system start-up, the meansigma methods of front 50 frame short-time energies is as the value of threshold value T；
(2), sound-content based on mel cepstrum coefficients and support vector machine is checked
Simple dependence volume detection is omitted child and is generally not present false dismissal, but false alarm rate is the highest, for reducing false alarm rate, at the beginning of volume
After inspection finds suspicious object, then use sound identification method to analyze sound-content, distinguish and whether sound exists cry, point
Cry, concretely comprises the following steps:
Step2.1: ask for logarithmic energy spectrum
Amplitude spectrum to t frame audio signalTake the logarithm, obtain logarithmic energy spectrum；
Step2.2: calculate mel cepstrum coefficients
The computing formula of mel cepstrum coefficients (MFCC) is:
Wherein,Representing the n-th mel cepstrum coefficients of t frame, K represents the number of Mel (Mel) bank of filters, at this
In invention, K takes 24, gives up flip-flop, and the present invention takes n=1, and 2 ..., 12；
Step2.3:SVM classifies
12 mel cepstrum coefficients are classified, if classification results is as feature, input support vector machines grader
1, then it is assumed that there is omission child, now start alarm device, alert information is sent to school bus director and higher level supervision department
Door, reminds related personnel to rescue in time.
The most according to claim 1 based on sob identification leave over child's detection method, it is characterised in that SVM classifier
Needing precondition to obtain, the training method of the present invention is: selects 100 sections and comprises the sound clip conduct that child cries and shout, screams
Positive sample, and 1000 sections of other sound clips are as negative sample, extract the mel cepstrum coefficients feature of Different categories of samples, select footpath
It is trained as kernel function to basic function, obtains SVM classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610314228.7A CN105976831A (en) | 2016-05-13 | 2016-05-13 | Lost child detection method based on cry recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610314228.7A CN105976831A (en) | 2016-05-13 | 2016-05-13 | Lost child detection method based on cry recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105976831A true CN105976831A (en) | 2016-09-28 |
Family
ID=56991780
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610314228.7A Pending CN105976831A (en) | 2016-05-13 | 2016-05-13 | Lost child detection method based on cry recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105976831A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107357292A (en) * | 2017-07-13 | 2017-11-17 | 上海斐讯数据通信技术有限公司 | Intelligent safeguard system and its maintaining method is seen in a kind of children's room |
| KR101807213B1 (en) * | 2017-05-18 | 2017-12-08 | 주식회사 보임 | Safety managing system for passenger of vehicle using speech recognition |
| CN108369813A (en) * | 2017-07-31 | 2018-08-03 | 深圳和而泰智能家居科技有限公司 | Specific sound recognition methods, equipment and storage medium |
| CN110390942A (en) * | 2019-06-28 | 2019-10-29 | 平安科技（深圳）有限公司 | Mood detection method and its device based on vagitus |
| WO2020151169A1 (en) * | 2019-01-23 | 2020-07-30 | 苏州美糯爱医疗科技有限公司 | Method for automatic removal of frictional sound interference of electronic stethoscope |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1835073A (en) * | 2006-04-20 | 2006-09-20 | 南京大学 | Mute detection method based on speech characteristic to jude |
| CN102426835A (en) * | 2011-08-30 | 2012-04-25 | 华南理工大学 | Method for identifying local discharge signals of switchboard based on support vector machine model |
| CN102664006A (en) * | 2012-04-14 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Abnormal voice detecting method based on time-domain and frequency-domain analysis |
| CN103730130A (en) * | 2013-12-20 | 2014-04-16 | 中国科学院深圳先进技术研究院 | Detection method and system for pathological voice |
| CN205034058U (en) * | 2015-09-10 | 2016-02-17 | 清华大学苏州汽车研究院（相城） | Infant alarm system in car that passes into silence |
-
2016
- 2016-05-13 CN CN201610314228.7A patent/CN105976831A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1835073A (en) * | 2006-04-20 | 2006-09-20 | 南京大学 | Mute detection method based on speech characteristic to jude |
| CN102426835A (en) * | 2011-08-30 | 2012-04-25 | 华南理工大学 | Method for identifying local discharge signals of switchboard based on support vector machine model |
| CN102664006A (en) * | 2012-04-14 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Abnormal voice detecting method based on time-domain and frequency-domain analysis |
| CN103730130A (en) * | 2013-12-20 | 2014-04-16 | 中国科学院深圳先进技术研究院 | Detection method and system for pathological voice |
| CN205034058U (en) * | 2015-09-10 | 2016-02-17 | 清华大学苏州汽车研究院（相城） | Infant alarm system in car that passes into silence |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101807213B1 (en) * | 2017-05-18 | 2017-12-08 | 주식회사 보임 | Safety managing system for passenger of vehicle using speech recognition |
| CN107357292A (en) * | 2017-07-13 | 2017-11-17 | 上海斐讯数据通信技术有限公司 | Intelligent safeguard system and its maintaining method is seen in a kind of children's room |
| CN108369813A (en) * | 2017-07-31 | 2018-08-03 | 深圳和而泰智能家居科技有限公司 | Specific sound recognition methods, equipment and storage medium |
| WO2019023877A1 (en) * | 2017-07-31 | 2019-02-07 | 深圳和而泰智能家居科技有限公司 | Specific sound recognition method and device, and storage medium |
| WO2020151169A1 (en) * | 2019-01-23 | 2020-07-30 | 苏州美糯爱医疗科技有限公司 | Method for automatic removal of frictional sound interference of electronic stethoscope |
| CN110390942A (en) * | 2019-06-28 | 2019-10-29 | 平安科技（深圳）有限公司 | Mood detection method and its device based on vagitus |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105976831A (en) | Lost child detection method based on cry recognition | |
| KR101969504B1 (en) | Sound event detection method using deep neural network and device using the method | |
| CN100485780C (en) | Quick audio-frequency separating method based on tonic frequency | |
| CN105809890B (en) | Towards omission child's detection method of safety of school bus | |
| KR101621774B1 (en) | Alcohol Analyzing Method, Recording Medium and Apparatus For Using the Same | |
| US20150045920A1 (en) | Audio signal processing apparatus and method, and monitoring system | |
| US9916844B2 (en) | Method for determining alcohol consumption, and recording medium and terminal for carrying out same | |
| US20150016617A1 (en) | Modified mel filter bank structure using spectral characteristics for sound analysis | |
| Xu et al. | ER: Early recognition of inattentive driving leveraging audio devices on smartphones | |
| KR20170120326A (en) | Method for providing sound detection information, apparatus detecting sound around vehicle, and vehicle including the same | |
| CN107554470B (en) | Apparatus and method for handling vehicle emergency status | |
| Turan et al. | Monitoring Infant's Emotional Cry in Domestic Environments Using the Capsule Network Architecture. | |
| Kandpal et al. | Classification of ground vehicles using acoustic signal processing and neural network classifier | |
| Harlow et al. | Automated accident detection system | |
| Mielke et al. | Smartphone application for automatic classification of environmental sound | |
| Kaur et al. | Traffic state detection using smartphone based acoustic sensing | |
| CN112420074A (en) | Method for diagnosing abnormal sound of motor of automobile rearview mirror | |
| KR101621780B1 (en) | Method fomethod for judgment of drinking using differential frequency energy, recording medium and device for performing the method | |
| Lieskovska et al. | Acoustic surveillance system for children’s emotion detection | |
| Noor et al. | Audio visual emotion recognition using cross correlation and wavelet packet domain features | |
| Balia et al. | A comparison of audio-based deep learning methods for detecting anomalous road events | |
| He et al. | Deep learning approach for audio signal classification and its application in fiber optic sensor security system | |
| Vij et al. | Transportation mode detection using cumulative acoustic sensing and analysis | |
| US9943260B2 (en) | Method for judgment of drinking using differential energy in time domain, recording medium and device for performing the method | |
| Ahmad et al. | Unsupervised multimodal VAD using sequential hierarchy |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160928 |