CN109938709A - The triggering information weight reselection procedure optimization system of wearable safety alarm device - Google Patents
The triggering information weight reselection procedure optimization system of wearable safety alarm device Download PDFInfo
- Publication number
- CN109938709A CN109938709A CN201910174132.9A CN201910174132A CN109938709A CN 109938709 A CN109938709 A CN 109938709A CN 201910174132 A CN201910174132 A CN 201910174132A CN 109938709 A CN109938709 A CN 109938709A
- Authority
- CN
- China
- Prior art keywords
- signal
- model
- identification model
- wearer
- optimization system
- 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
Abstract
The present invention relates to the alerting techniques of wearable device, the triggering information weight reselection procedure optimization system of specially wearable safety alarm device, including signal identification model, decision model and the alarm modules being connected；Wherein signal identification model is used to receive and analyze the physiological signal and environmental signal of extraneous offer, judges the ambient condition and oneself state of wearer；Decision model is used for the ambient condition and oneself state calculated according to signal identification model, carries out different weight distributions to signal identification result and provides the scoring of weighted sum；Alarm modules are used to compare the scoring and preset alarm threshold value that decision model provides, and judge whether to trigger and sound an alarm.The system decides whether triggering warning device for the analysis result of different surrounding enviroment signals, Model Weight accounting, improves the efficiency remotely detected to wearer's personal safety.
Description
Technical field
The present invention relates to the alerting techniques of wearable device, the touching of algorithm level in specially wearable safety alarm device
Photos and sending messages weight selects optimization system.
Background technique
The disadvantaged group of children and old man as society, personal safety is a focus of social concerns always.Simultaneously
It individually goes out also generate worry for old parents.In on ordinary days, when children or old man not at one's side when, can be used can wear
Equipment is worn to detect its personal safety.Existing wearable safety watch is based on less detection signal, and efficiency, precisely
Degree aspect is to be improved, cannot accurately detect the situation of children.
Summary of the invention
In order to improve the efficiency and accuracy that wearable device determines wearer's personal safety, the present invention proposes wearable
The triggering information weight reselection procedure optimization system of safety alarm device, for different surrounding enviroment signals, Model Weight accounting
Analysis result decides whether triggering warning device, improves the efficiency remotely detected to wearer's personal safety.
The technical scheme adopted by the invention is that: the triggering information weight reselection procedure of wearable safety alarm device optimizes system
System, including signal identification model, decision model and the alarm modules being connected；Wherein signal identification model is for receiving and analyzing
The physiological signal and environmental signal that the external world provides, judge the ambient condition and oneself state of wearer；Decision model is used for basis
The ambient condition and oneself state that signal identification model calculates carry out different weight distributions to signal identification result and provide weighting
The scoring of summation；Alarm modules are used to compare the scoring and preset alarm threshold value that decision model provides, and judge whether that triggering is concurrent
Alarm out.
Preferably, physiological signal includes pulse wave signal, blood pressure signal and skin electric signal, and environmental signal includes voice signal
And acceleration signal；Signal identification model includes that Emotion identification model, tone identification model, the keyword based on physiological data are known
Other model, Emotion identification model, specific sound identification model, gesture recognition model, moving state identification model are based on physiology number
According to Emotion identification model judge according to physiological signal the oneself state of wearer, tone identification model, keyword recognition model,
Emotion identification model, specific sound identification model judge the oneself state and environmental signal of wearer, appearance all in accordance with voice signal
State identification model, moving state identification model judge the oneself state of wearer according to acceleration signal.
Compared with prior art, the present invention achieves following technical effect:
The present invention judges that wearer is by the acquisition to wearer's physiological signal, voice signal, acceleration signal information
It is no to be in a kind of improper circumstances, and then decide whether to carry out remote alarm；Its specific optimal way has with local environment
Close, according to varying environment (such as strenuous exercise, noisy environment), signal contribution degree, two aspects of identification model contribution degree into
The distribution of row weight, and then acquisition information is made to be utilized effectively, accuracy is improved, to achieve the purpose that intelligent alarms.It can
It is widely used in wearable device (wrist-watch, bracelet), improves the efficiency remotely detected to wearer's personal safety.
Detailed description of the invention
Fig. 1 is system structure diagram of the invention.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Embodiment
Referring to Fig. 1, in the present embodiment, the triggering information weight reselection procedure optimization system of wearable safety alarm device includes phase
Signal identification model, decision model and the alarm modules of connection；Wherein signal identification model is for receiving and analyzing extraneous offer
Physiological signal and environmental signal, judge the ambient condition and oneself state of wearer；Decision model is used for according to signal identification
The ambient condition and oneself state that model calculates, carry out different weight distributions to signal identification result and provide weighted sum to comment
Point；Alarm modules are used to compare the scoring and preset alarm threshold value that decision model provides, and judge whether to trigger and sound an alarm.Its
In, physiological signal includes pulse wave signal, blood pressure signal and skin electric signal, and environmental signal includes voice signal and acceleration letter
Number.
Signal identification model includes Emotion identification model, tone identification model, keyword recognition mould based on physiological data
Type, Emotion identification model, specific sound identification model, gesture recognition model, moving state identification model, based on physiological data
Emotion identification model judges the oneself state of wearer, tone identification model, keyword recognition model, mood according to physiological signal
Identification model, specific sound identification model judge the oneself state and environmental signal of wearer all in accordance with voice signal, and posture is known
Other model, moving state identification model judge the oneself state of wearer according to acceleration signal.This six kinds of models are using warp
The machine learning algorithm of initial acquisition data set training judges the probability of data fit specified conditions；It is secondary when being judged by accident in use process
When number is more than threshold value, it will use data that initial acquisition data set and Retraining algorithm is added.
Emotion identification model based on physiological data uses training data training machine learning algorithm, and mood is classified as two
Class: smaller negative emotions, neutral mood and all positive moods are classified as one kind, and more negative emotions are classified as another kind of；
Trained model is called when use, judges whether wearer has by analysis pulse wave signal, blood pressure signal and skin electric signal
Larger negative emotions.
Tone identification model exports according to voice signal and has fear in wearer and short distance in the tone of talker
The probability of the property tone and the abusing property tone；Keyword recognition model exports wherein according to voice signal comprising emergency property words and disgrace
The probability of scolding property words；Emotion identification model exports the probability wherein comprising frightened mood according to voice signal；Specific sound is known
Other model exports probability wherein comprising preset specific sound according to voice signal.
Gesture recognition model judges that wearer is in the violent shake, defensive of hand struggle according to acceleration signal
Pat, fall down with acceleration it is excessive or mutation appearance probability of state；It moves identification model and wearer is judged according to acceleration signal
Body movement situation, and export movement severity.
Decision model carries out each data reliability judgement to the result that signal identification model exports, according to varying environment state
Under different weight ratios, scoring is weighted to each result, the scoring after output summation.Power in decision model under varying environment
It again can be by initial data and using the new raw data set of Data Synthesis and to algorithm than after accumulation reaches certain erroneous judgement number
Re -training is carried out, the data weighting ratio under varying environment is updated.
When wearer is in strenuous exercise's state, mesh can be characterized by the collected acceleration signal of three-axis sensor
Preceding wearable device is just in strenuous exercise.The weight distribution algorithm pre-set is called at this time, and weight point is carried out to unlike signal
With processing.Pulse wave signal substantially changes similar, the feelings to the pulse wave signal when mood swing when due to strenuous exercise
The mood result obtained under condition by pulse wave signal is by larger interference, therefore pulse wave signal weight distribution can be transferred to low water
Flat, voice signal (i.e. voice signal) can play a leading role (specific weight value is provided by many experiments analysis) at this time.Other letters
Number and identification model analysis result weight distribution it is similar with its.It determines eventually by the analysis result of weight accounting under the environment
It is fixed whether to trigger warning device.
It is special locating for equipment that the present invention is carried out by the signal (voice signal, acceleration signal) being associated with the external world first
The judgement of environment, according to voice signal obtain facility environment noisy degree, acceleration signal analyze wearer motion state and
Posture, identifies the movement of wearer, and then judges the state of wearer itself；It is called according to the ambient condition of wearable device different
Weight distribution algorithm assigns different weighted values to the contribution degree of various signals, separates main signal and secondary signal, carry out thereafter
Analysis；Different weight distribution algorithms are called according to the ambient condition of wearable device, weight is carried out to each signal identification model
Distribution, i.e., the identification model result be " effective " when, to warning device wake-up percentage contribution score, only score value is total
With warning device can be just triggered when reaching certain numerical value.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (8)
1. the triggering information weight reselection procedure optimization system of wearable safety alarm device, which is characterized in that including the letter being connected
Number identification model, decision model and alarm modules；Wherein signal identification model is used to receive and analyze the physiology letter of extraneous offer
Number and environmental signal, judge the ambient condition and oneself state of wearer；Decision model is used to be calculated according to signal identification model
Ambient condition and oneself state, different weight distributions are carried out to signal identification result and provide the scoring of weighted sum；Alarm
Module is used to compare the scoring and preset alarm threshold value that decision model provides, and judges whether to trigger and sound an alarm.
2. triggering information weight reselection procedure optimization system according to claim 1, which is characterized in that physiological signal includes pulse
Wave signal, blood pressure signal and skin electric signal, environmental signal include voice signal and acceleration signal；
Signal identification model includes Emotion identification model, tone identification model, keyword recognition model, feelings based on physiological data
Thread identification model, specific sound identification model, gesture recognition model, moving state identification model, the mood based on physiological data
Identification model judges the oneself state of wearer, tone identification model, keyword recognition model, Emotion identification according to physiological signal
Model, specific sound identification model judge the oneself state and environmental signal of wearer, gesture recognition mould all in accordance with voice signal
Type, moving state identification model judge the oneself state of wearer according to acceleration signal.
3. triggering information weight reselection procedure optimization system according to claim 2, which is characterized in that the feelings based on physiological data
Thread identification model uses training data training machine learning algorithm, and mood is classified as two classes: smaller negative emotions, neutral mood
And all positive moods are classified as one kind, more negative emotions are classified as another kind of；Trained model is called when use, is led to
It crosses analysis pulse wave signal, blood pressure signal and skin electric signal and judges whether wearer has larger negative emotions.
4. triggering information weight reselection procedure optimization system according to claim 2, which is characterized in that tone identification model according to
Voice signal exports the probability for having the phobia tone and the abusing property tone in wearer and short distance in the tone of talker；
Keyword recognition model exports the probability wherein comprising emergency property words and abusing property words according to voice signal；Emotion identification mould
Type exports the probability wherein comprising frightened mood according to voice signal；Specific sound identification model exports wherein according to voice signal
Probability comprising preset specific sound.
5. triggering information weight reselection procedure optimization system according to claim 2, which is characterized in that gesture recognition model according to
Acceleration signal judge wearer be in the violent shake of hand struggle, defensive beating, fall down with acceleration it is excessive or
The appearance probability of state of mutation；Movement identification model judges the body movement situation of wearer according to acceleration signal, and exports fortune
Dynamic severity.
6. triggering information weight reselection procedure optimization system according to claim 1, which is characterized in that decision model knows signal
The result of other model output carries out each data reliability judgement, according to the different weight ratios under varying environment state, to each result
It is weighted scoring, the scoring after output summation.
7. triggering information weight reselection procedure optimization system according to claim 1, which is characterized in that different rings in decision model
Weight ratio under border can be by initial data and the raw data set new using Data Synthesis after accumulation reaches certain erroneous judgement number
And re -training is carried out to algorithm, update the data weighting ratio under varying environment.
8. triggering information weight reselection procedure optimization system according to claim 2, which is characterized in that each in signal identification model
Model uses the probability that data fit specified conditions are judged through the machine learning algorithm of initial acquisition data set training；When using
When judging number by accident more than threshold value in journey, it will use data that initial acquisition data set and Retraining algorithm is added.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910174132.9A CN109938709A (en) | 2019-03-08 | 2019-03-08 | The triggering information weight reselection procedure optimization system of wearable safety alarm device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910174132.9A CN109938709A (en) | 2019-03-08 | 2019-03-08 | The triggering information weight reselection procedure optimization system of wearable safety alarm device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109938709A true CN109938709A (en) | 2019-06-28 |
Family
ID=67009331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910174132.9A Pending CN109938709A (en) | 2019-03-08 | 2019-03-08 | The triggering information weight reselection procedure optimization system of wearable safety alarm device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109938709A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210592A (en) * | 2020-01-07 | 2020-05-29 | 珠海爬山虎科技有限公司 | Video identification monitoring method, computer device and computer readable storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109407504A (en) * | 2018-11-30 | 2019-03-01 | 华南理工大学 | A kind of personal safety detection system and method based on smartwatch |
-
2019
- 2019-03-08 CN CN201910174132.9A patent/CN109938709A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109407504A (en) * | 2018-11-30 | 2019-03-01 | 华南理工大学 | A kind of personal safety detection system and method based on smartwatch |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210592A (en) * | 2020-01-07 | 2020-05-29 | 珠海爬山虎科技有限公司 | Video identification monitoring method, computer device and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109407504B (en) | Personal safety detection system and method based on smart watch | |
| Wang et al. | Fall detection based on dual-channel feature integration | |
| Wang et al. | Walking pattern classification and walking distance estimation algorithms using gait phase information | |
| Mannini et al. | Accelerometry-based classification of human activities using Markov modeling | |
| CN107773214A (en) | A kind of method, computer-readable medium and the system of optimal wake-up strategy | |
| US20210064141A1 (en) | System for detecting a signal body gesture and method for training the system | |
| CN104468981B (en) | A kind of campus despot based on intelligent mobile phone platform insults detection method | |
| CN105303183B (en) | A kind of child posture discriminance analysis system and method based on wearable device | |
| Xefteris et al. | Performance, challenges, and limitations in multimodal fall detection systems: A review | |
| Carfi et al. | Online human gesture recognition using recurrent neural networks and wearable sensors | |
| KR20190134911A (en) | Electronic device and method for controlling the electronic device thereof | |
| Ding et al. | Energy efficient human activity recognition using wearable sensors | |
| CN108010273A (en) | A kind of drowned detection alarm based on Cloud Server and machine learning algorithm | |
| CN109938709A (en) | The triggering information weight reselection procedure optimization system of wearable safety alarm device | |
| CN113397494A (en) | Animal sign monitoring device and method and intelligent wearable device | |
| CN110807471B (en) | Behavior recognition system and recognition method of multi-mode sensor | |
| CN105551191B (en) | A kind of fall detection method | |
| Zheng et al. | L-sign: Large-vocabulary sign gestures recognition system | |
| Ning et al. | Fall detection algorithm based on gradient boosting decision tree | |
| Tay et al. | A review of abnormal behavior detection in activities of daily living | |
| Lin et al. | Adaptive multi-modal fusion framework for activity monitoring of people with mobility disability | |
| Qu et al. | Convolutional neural network for human behavior recognition based on smart bracelet | |
| Xie et al. | ART: adaptive and real-time fall detection using COTS smart watch | |
| CN111626273A (en) | Fall behavior recognition system and method based on atomic action time sequence characteristics | |
| Smailov et al. | A Novel Deep CNN-RNN Approach for Real-time Impulsive Sound Detection to Detect Dangerous Events |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190628 |