CN106338935A - Robot emotion recognition method and system - Google Patents
Robot emotion recognition method and system Download PDFInfo
- Publication number
- CN106338935A CN106338935A CN201510198617.3A CN201510198617A CN106338935A CN 106338935 A CN106338935 A CN 106338935A CN 201510198617 A CN201510198617 A CN 201510198617A CN 106338935 A CN106338935 A CN 106338935A
- Authority
- CN
- China
- Prior art keywords
- electro
- emotion recognition
- emotion
- physiological signals
- module
- 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
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
- G05B19/042—Programme control other than numerical control, i.e. in sequence controllers or logic controllers using digital processors
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/25—Pc structure of the system
- G05B2219/25252—Microprocessor
Abstract
The invention provides a robot emotion recognition method and system. The robot emotion recognition method comprises the steps of detecting physiological electric signals of a human body; extracting a characteristic waveform of the physiological electric signals; performing physiological abnormality detection according to the extracted characteristic waveform; and performing emotion recognition according to the extracted characteristic waveform. According to the invention, physiological electricity is acquired at different emotional states, necessary physiological electricity characteristics are extracted, an emotional physiological electricity model is trained according to an extracted characteristic set, the current emotional state of a wearer can be judged by using the model and the current physiological electric signals, a heart rate abnormality result and an emotional state judgment result are managed by using a mobile phone App (Application), and the wearer is helped to know physical and psychological health conditions thereof in real time. In addition, devices which are inconvenient to wear such as a camera in the traditional method are abandoned, and the portability is improved.
Description
Technical field
The present invention relates to artificial intelligence field is and in particular to a kind of robot emotion identification method and system.
Background technology
The destruction of the development of society and natural environment leads to sub-health population and chronic organic sick people fast
Speed increases, and most of organic disease such as essential hypertension, gastric ulcer, cardiovascular and cerebrovascular disease, cancer etc.
All there is very big relevance with personal mood；Additionally, with the intensification further of China human mortality aging, empty nest
Old man and old solitary people population are also in rising trend.For the above social situations of China, it is directed in the urgent need to one kind
Inferior health under pressure, organic disease and the psychological consolation of mid-aged population and the solution of motion management, mesh
Before, amateur medical small physiological electricity custodial care facility emerges in an endless stream, and its main target is to aid in family health care and enters
Row self-monitoring and management, the technological means of employing is mainly with or makes single equipment, or collects successfully
Mobile phone can be arrived, carry out physiology electric collection and display, and in the case of necessary by letter is carried out to such as electrocardiosignal
Single process to calculate the real-time heart rate of human body, and extremely carries out suitable health of heart early warning according to heart rate, this kind of sets
The standby pre- diagnosis for personal health management and doctor to a certain extent provides conveniently, but it is only opposite
Reason electric signal is shown, only starts with although can be to one on the preferable electrocardio equipment processing from analysis heart rate
Heart rate unusual condition relatively common a bit be given some prompting, but electro-physiological signals can not be done emotion recognition with
And build intelligent robot feedback system；Additionally, in emotional robot field, major part can carry out emotion
The technology of interaction is substantially based on image recognition and speech recognition technology, and the use of this kind of technology typically requires borrows
Help shooting first-class this kind of can not realize wearable complicated peripheral hardware, not readily portable and safeguard, and due to expressing one's feelings
Can be controlled by nerve, the usual error of emotion emotional state of expression reaction is very big.
Content of the invention
The present invention is directed to the above-mentioned problems in the prior art, proposes a kind of robot emotion recognition scheme, should
Scheme by being monitored and analyzed to physiology electric, and based on analysis result provide intelligentized emotion recognition and
Motion management service plan；It uses small physiology electrical sensing piece to replace traditional camera scheme simultaneously, can
Realize conveniently dressing or wearing.
The technical scheme is that
A kind of robot emotion identification method, including step:
The electro-physiological signals of detection human body；
Extract the signature waveform of described electro-physiological signals；
Described signature waveform according to being extracted carries out physically different detection；
Using lda algorithm, emotion recognition is carried out to the described signature waveform being extracted.
Preferably, also include before the characteristic data step of the described electro-physiological signals of described extraction:
Remove the unavailable data in described electro-physiological signals；
Electro-physiological signals described in normalized；
LPF is carried out to described electro-physiological signals.
Preferably, the described characteristic extracting described electro-physiological signals method particularly includes:
Extract the pqrst signature waveform of human ecg signal.
Preferably, described physically different detection is carried out according to the described characteristic extracted method particularly includes:
Using the inverse of phase between r popin is equal in certain time as heart rate；
Physically different detection is carried out according to described heart rate.
Preferably, described use lda algorithm carries out the concrete of emotion recognition to the described signature waveform being extracted
Method is:
Gather the signature waveform sample matrix x under variant emotion and set up emotional category vector v, described
The vector that emotional category vector is formed by variant emotion identification；
Pick out the sample matrix x of each classification according to described v respectively, calculate the sample of variant classification
This within class scatter matrix and total inter _ class relationship matrix:
..., k
Wherein, x is the sample data matrix of input, and behavior sample one by one is classified as feature one by one；It is all kinds of
Sample average；K is sample generic number；
Calculate scatter matrix between sample:
Wherein, it is Different categories of samples number, be sample overall mean；
Take the characteristic vector composition transition matrix w corresponding to front k-1 characteristic value of matrix；
Central point after formula w* is calculated each class projection；
New according to the far and near judgement of Euclidean distance of new feature samples all kinds of central point of distance after matrix w projection
The affiliated emotional category of sample data.
The present invention also provides a kind of robot emotion recognition system, sets including separate signal detection front end
Standby and mobile terminal device, described signal detection headend equipment and mobile terminal device carry out wireless communication connection,
Described signal detection headend equipment is used for gathering and pre-processing Human Physiology electric signal and send to described movement
Terminal device, described mobile terminal device is used for being carried out according to the described pretreated electro-physiological signals being received
Physically different detection and emotion recognition.
Preferably, described signal detection headend equipment includes being sequentially connected signal collecting device, multiplying arrangement,
Filter apparatus, ad conversion equipment and main control device, described signal collecting device is used for gathering electro-physiological signals,
Described multiplying arrangement is used for amplifying the electro-physiological signals of described collection, and described filter apparatus are used for the life after amplifying
Reason electric signal carries out clutter and filters, and described ad conversion equipment is used for filtered simulation electro-physiological signals conversion
For data signal and be sent to described main control device, it is simultaneously that it is wireless that described main control device receives described data signal
Send to described mobile terminal device.
Preferably, the power management that described signal detection headend equipment also includes being connected with described main control device sets
Standby, for power supply power consumption management is carried out to described signal detection headend equipment.
Preferably, described mobile terminal device includes middle control module and connected signal characteristic abstraction mould
Block, physically different detection module and emotion recognition module, described signal characteristic abstraction module is used for extracting described
The signature waveform of electro-physiological signals, described physically different detection module is used for being carried out according to the signature waveform being extracted
Physically different detection, described emotion recognition module is used for carrying out emotion recognition according to the signature waveform being extracted.
Preferably, described mobile terminal device also includes judge module and the health official communication being connected with described middle control module
Ask module, described judge module is used for judging physically different testing result and emotion recognition result, described health is consulted
Ask module to be used for when described judge module judges physically different or produce during the disturbance of emotion, providing health to consult
Ask and inform and carry out physically different detection or emotion recognition again, otherwise inform that user's testing result is normal and exits
Detection, the described disturbance of emotion is Negative Affect.
The invention has the following beneficial effects:
1st, adopt the physiology electro-detection front end of complete or collected works' accepted way of doing sth, making an uproar of generation can be placed in motion or remote electrode
Carry under acoustic environment
Take, put big and filter faint bioelectrical signals, filter output signal setup time is short, high cmrr,
And power consumption is extremely low, small volume.
2nd, by power supply rational on software and power managed so that whole system can rely on a button electricity
Pond or compact lithium cell are powered and are worked several weeks, substantially reduce with respect to traditional physiology electric patient monitor cost, measurement
Simply, portable it is adaptable to domestic medicine application and wearable application.
3rd, cooperation Low Power Embedded Microprocessor collection electro-physiological signals, and sent out by low-power consumption bluetooth technology
Deliver in terminals such as mechanical, electrical brains, be easy to gather, signal quality is higher, and system power dissipation is low.
4th, health consultation service is provided the user by mobile terminals such as mobile phones, additionally need not increase other equipment,
Meet the demand to simplification and convenience for the modern electronic product user.
Brief description
In order to be illustrated more clearly that the technical scheme of the embodiment of the present invention, below will be to required in embodiment description
Accompanying drawing do simple introduce it should be apparent that, drawings discussed below is only some embodiments of the present invention,
For those of ordinary skill in the art, on the premise of not paying creative work, can also be according to these
Accompanying drawing obtains other accompanying drawings.
Fig. 1 is the flow chart of robot of the present invention emotion identification method.
Fig. 2 is the structured flowchart of the robot emotion recognition system of the present invention.
Fig. 3 is the flow chart of the method for managing power supply of the present invention.
Fig. 4 is the topology diagram of low pass filter in the present invention.
Fig. 5 is the topology diagram of bivalent high-pass filter in the present invention.
Specific embodiment
With reference to Figure of description, the specific embodiment of the embodiment of the present invention is elaborated.
As shown in figure 1, robot of the present invention emotion identification method includes step:
S101: the electro-physiological signals of detection human body；
Gather under different emotional states (tranquil, glad, sad, frightened, indignation, nausea, surprised)
Physiological electricity of human body, for ease of description, the unified electrocardiosignal for human body of the electro-physiological signals in present embodiment,
Collection to electrocardiosignal can be realized by two induction electrode e1 and e2 leading, and e1, e2 lead for two
Electrode input, is connected with the Self-adhesive electrode slice being attached to body skin surface, the corresponding electrode of wherein e1
Piece is pasted on underbelly lower left, and the corresponding electrode slice of e2 is pasted on above heart.
The electrocardiosignal collecting is usually present many noises, and it is right to need before carrying out further signal transacting
Signal carries out the pretreatment of denoising, the heart being adopted according to the feature of this product hardware output signal, present embodiment
Electric signal preprocess method comprises each link described in following step s102 to s104, mainly using ad8232
Chip realizes pretreatment.
S102: remove the unavailable data in described electro-physiological signals；
It has been generally acknowledged that electrocardiosignal is by the cyclical signal of pqrst wave component, and due to motion or electrode
Shakiness would generally cause the aperiodicity of signal, therefore, here using checking whether certain segment signal is periodic signal
To judge whether electrocardiosignal receives the larger pollution of ratio, to decide whether subsequently to be located with this segment signal
Reason.Ad8232 is mainly adopted as physiology electro-detection front end, for the electrocardio detecting in present embodiment
Signal is pre-processed, and described induction electrode e1 and e2 is connected to described ad8232, and described ad8232 is
The signal condition chip of one complete or collected works' accepted way of doing sth, in the case of having motion or the noise that remotely prevents
Extract, amplify and filter faint bioelectrical signals.Ad8232 adopts 4mm × 4mm, no pin
Lfcsp-20 encapsulates, and volume is very little, its built-in instrument amplifier, operational amplifier, one
Right leg drive amplifier and an intermediate power supplies voltage reference voltage buffer, additionally built-in lead-fail detector detection
Circuit and an automatic taibiter, rapid after reconnect of leading can recover signal, reduce high pass filter
Ripple device long-tail setup time.
S103: electro-physiological signals described in normalized；
Electrocardiosignal due to different people may have the larger specific difference of ratio in amplitude, unites for convenience
One process, needs electrocardiosignal all to unify to incorporate on identical yardstick, and this method is referred to as normalizing.Right
Through being judged to that available one piece of data is normalized, using following mathematical formulae:
Wherein, be this segment data each point, be the maximum of this segment data, be the minimum of a value of this segment data.
S104: described electro-physiological signals are carried out with LPF:
The useful information frequency of electrocardiosignal is typically located at 3-40hz, and other frequency contents are largely made an uproar
Sound introduce it is therefore desirable to these noise frequencies give remove.Described ad8232 can be by the choosing of peripheral components
Select configuration one two limit high-pass filter and three limits low pass filter, to eliminate correction of motion artefacts and
Extra high-frequency noise.
LPF aspect adopts the topological structure of sallen-key wave filter, as shown in figure 4, passing through
Three 1m resistance outside opamp+, refout, opamp- and iaout pin, a 100k electricity
Resistance, a 10nf electric capacity and 1.5nf electric capacity one second-order low-pass filter of composition, cut-off frequency 41.1hz,
Gain 11.There is maximally-flat degree and sharpen cut-off frequency.
Described ad8232 passes through outer two 10m of hpdrive, iaout, sw and refout pin
Resistance, 1.4m resistance, 1uf electric capacity, a 10uf electric capacity, may make up a second order high pass
Wave filter, is used for filtering noise.As shown in figure 5, cut-off frequency 0.05hz, gain 100.This circuit topology
Less cut-off frequency can be provided using less resistance capacitance value, and there is band logical the most flat, low
Frequency suppresses preferably, and output impedance is very low.
Electrocardiosignal with noise and clutter is filtered by high-pass filter and low pass filter jointly, carries
Take out the required effective electrocardiosignal with human body emotional information, 0.05hz-41.1hz meets Algorithm Analysis
Required data.
S105: extract the signature waveform of described electro-physiological signals；
Correctly to identify electrocardiosignal it is necessary to extract the signature waveform of electrocardiosignal, mainly pqrst ripple.
Present embodiment adopts the method for wavelet transformation to extract signature waveform, and concrete grammar is as follows:
1) three layer scattering wavelet transformations (mallat algorithm) are done to one section of electrocardiosignal after denoising and digitlization,
The conversion signal length obtaining each layer is equal with original signal strength, and 20 points before and after the signal after conversion are put
It is zero, then finds out pole sizes values point, point herein is bigger (or little) than former point or latter point value, then will
Its sequence, finds out 8 values of maximum, asks it to be averagely worth to thrmax.By the signal sequence after conversion
Afterwards, find out front 100 values of minimum, take its mean value zerovalue.
2) threshold value selects and positioning r ripple
Value is set to thr=(thrmax-zerovalue) * 30%.
The maximum of acquirement is compared with threshold value thr, is exactly the value of r wave point if greater than threshold value.
3) calibrate the extraction accuracy (exclusion flase drop) of r ripple
If two adjacent maximum points be smaller than 0.4, less one of amplitude of removing.
4) other ripples are detected according to r ripple, mainly the pitch of waves according to r ripple and other waveforms, in certain scope
Inside take maximum and minimum of a value.
For q ripple and s ripple, minimum of a value is taken to can get in 10 points after r wavefront respectively；For p
Ripple and t ripple, take maximum and s between q wavefront 0.02*sfreq-0.4*rrmean*sfreq respectively
Maximum between 0.02*sfreq-0.4*rrmean*sfreq after ripple.(sfreq is sample rate,
Rrmean is the mean value of phase between this segment signal rr)
S106: the described signature waveform according to being extracted carries out physically different detection；
In order to quickly calculate heart rate, present embodiment calculate 5s electrocardiogram (ECG) data process the r popin obtaining equal between the phase,
Using the inverse of this phase as heart rate；Continuously measured using sliding window technology, resolution ratio is 1s.
Carry out the detection of common several heart rate abnormal conditions in terms of heart rate abnormality detection, including mistake aroused in interest
Speed, areocardia and VPB.Judged according to heart rate: if the average heart rate in 5s is more than 110,
Then think and currently belong to tachycardia；If the average heart rate in 5s is less than 55 then it is assumed that currently belonging to the heart
Fight low；If the r wave number that between maximum r ripple, the phase is more than in 240, or 5s in 5s is less than 5, then recognize
For currently belonging to VPB, other situations belong to normally.These judge that data is all according to experiment and doctor
Experience determining.
S107: the described signature waveform according to being extracted carries out emotion recognition；
The statistical nature in the 60s time such as the amplitude of calculating pqrst ripple, a phase, power, several respectively
Examples of features is as follows:
Ecgp-std: electrocardio p ripple variance, that is, in each cycle p ripple variance；
Ecgpampl-range: the range difference of electrocardio p wave amplitude, i.e. electrocardio p wave amplitude in each cycle
The difference of maxima and minima；
Ecgpq-range: electrocardio pq wave amplitude away from range difference, that is, in each cycle p ripple and q wave amplitude away from
The difference of maxima and minima；
Ecgabsdiff γ ampl-mean: the intermediate value of electrocardio adjacent t-t wave amplitude absolute difference；
The variance of ecgdifframpl-std: electrocardio adjacent r-r wave amplitude difference；
Rateuptime_min: the minimum of a value of time used by heart rate rising edge；
Rateupr_max: the maximum of heart rate rate of rise size；
Rateamp-std: the variance of heart rate amplitude；
Ratesteptime_max: the maximum of heart rate rise time；
Classified using lda (analysis of linear discriminant analysis linear discriminent) method, base
This thought is high dimensional data to be projected on the lower dimensional space being easiest to classify, and basic procedure is:
1) such as: for every one piece of data, we detect pqrst ripple using method noted earlier, and calculate
As above sample characteristics.1467 segment datas can altogether be gathered, then obtain a sample matrix x, Yi Jiyi
Individual vector v, only comprises 0,1,2,3,4,5,6 this 7 numerals in this vector, represent 7 kinds of emotional states respectively
(0- ＞ is tranquil；1- ＞ is glad；2- ＞ is sad；3- ＞ is frightened；4- ＞ indignation；5- ＞ detests；6- ＞ is surprised), i.e. 7 kinds of classifications.This
7 numerals identify the emotional state represented by each sample of sample matrix x one by one；
2) pick out the sample matrix of each classification according to v respectively, be calculated every class sample using equation below
This within-cluster variance
Matrix si(represent class in extent of polymerization) and they add and sw(represent the poly- of sample generally of all categories
Conjunction degree):
7 scatter matrixes and 1 total inter _ class relationship matrix in sample class are calculated according to sample matrix x:
Wherein, x is the sample data matrix of input, and behavior sample one by one is classified as feature one by one；It is all kinds of
Sample average；K is sample generic number, k=7；
3) scatter matrix between sample is calculated according to below equation: (represent Different categories of samples journey discrete each other
Degree, you can discrimination)
Wherein, it is Different categories of samples number, be sample overall mean；
4) take the characteristic vector composition transition matrix w corresponding to front 6 characteristic values of matrix；
5) central point after formula w* is calculated each class projection；
6) the Euclidean distance new sample of far and near judgement according to new feature samples all kinds of central point of distance after matrix w projection
Originally which kind of belongs to, that is, new samples belong to that class emotion in all kinds of central point Euclidean distances nearest.Then
It may determine that current electrocardiogram (ECG) data segment table levies belonged to which kind of mood.Mood in the personal test of each training
In identification, the method recognition accuracy that the present invention adopts is up to more than 90%.
As shown in Fig. 2 the present invention also provides a kind of robot emotion recognition system 200, including separate
Signal detection headend equipment 210 and mobile terminal device 220, described signal detection front end 210 produces concrete
When its outward appearance can be made the wearable pattern such as wrist-watch, bangle, in order to carry and prevent lose, described movement
Terminal device 220 is mobile phone or notebook computer etc..Described signal detection headend equipment 210 and mobile terminal set
Standby 220 carry out wireless communication connection, and present embodiment carries out radio communication using low-power consumption bluetooth module, tradition
Bluetooth (bluetooth 2.0, bluetooth 2.1, bluetooth 3.0) technology power consumption is larger, be unfavorable for wearable device long when
Between data transfer.And the time of matching longer it is difficult to lifting user experience.Table 1 from transmission range, speed,
In the parameters such as response time, energy consumption, conventional Bluetooth and low-power consumption bluetooth (bluetooth 4.0) are contrasted:
Table 1
As can be seen that low-power consumption bluetooth has very big advantage from the contrast of trade mark, the present invention adopts low-power consumption
The physiology electric data is activation that described signal detection headend equipment 210 is collected by bluetooth is to stating mobile terminal device
220, in such as mobile phone app or computer, carry out according to being received in order to state mobile terminal device 220
Physiology electric data carries out physically different detection and emotion recognition.
Described signal detection headend equipment 210 includes signal collecting device 211, the multiplying arrangement being sequentially connected
212nd, filter apparatus 213, ad conversion equipment 214 and main control device 215, described signal collecting device 211,
Multiplying arrangement 212, filter apparatus 213, ad conversion equipment 214 can use induction electrode and ad8232 core
Piece is unified to be realized, and described main control device 215 selects the cc2541f256 chip of ti, and it is integrated with 2.4ghz
The on-chip system (object wearing device) of low-power consumption bluetooth (ble), transmitting receive electric current be respectively 18.2ma and
17.9ma, has three kinds of low-power consumption standby patterns, and minimum current only has 0.5ua.
Described cc2541f256 adopts the no pin qfn-40 encapsulation of 6mm × 6mm, a built-in high property
Can 8051 kernels, the flash of 256kb, 8kb obtain ram, also include Five-channel dma, three fixed
When device, 8 Channel 12-Bit adc, two uart, i2The peripheral hardwares such as c interface, coordinate 32mhz external crystal-controlled oscillation
Work, fully meets performance and the power consumption requirements of the present invention.
It is sent to described cc2541f256 core after described ad8232 chip carries out ECG signal processing
Piece, the electrocardiosignal being received is sent to by described cc2541f256 by its own integrated low-power consumption bluetooth
Mobile phone app or computer.
As preferred embodiment, described signal detection headend equipment 210 also includes and described main control device
The power supply management device 216 of 215 connections, for carrying out power supply to entirely described signal detection headend equipment 210
Power managed.From tps78225 as voltage stabilizing chip, this is a low pressure difference linear voltage regulator, and input comes
From lithium battery, input voltage 2.5-5v, meet the cell voltage 2.7v-4.2v scope of lithium battery, output voltage
2.5v, output current can reach more than 230ma.Cc2541 chip is opened by button
P-mosfet s12305, the Enable Pin of output high level to tps78225, can be on hardware
Cc2541 and whole circuit are powered, and cut-off current is minimum, and in shutdown, leakage current only has 18na.When lithium electricity
Cell voltage as little as 2.7v about when, that is, tps78225 input and output pressure reduction only have 200mv when remain to export
The electric current of more than 150ma, fully meets application needs, substantially make use of the electricity of lithium battery.In addition,
Described lithium battery voltage is input to cc2541 chip analog-digital converter after passing through electric resistance partial pressure carries out battery electricity
Amount monitoring.
When being charged to described lithium battery, from bq24080 as charging management chip, input is logical
Cross hard contact and access 5v voltage, export 4.2v to lithium cell charging.Can be shown by red led etc.
Show charged state, quick flashing represents and charges, lamp goes out, and expression is uncharged, and often bright expression has been filled with.Lithium electricity simultaneously
Pond external harmoniousness battery protecting plate, can prevent over-charging of battery or cross from putting, extend battery life.
Power managed:
Main control chip cc2541 provides three kinds of different operational modes: aggressive mode, idle pulley and low-power consumption
Pattern pm1, pm2 and pm3.The realization that super low-power consumption runs is passed through closing power module and avoids static state to let out
Dew power consumption, reduces dynamic power consumption also by using gated clock and closing oscillator.Several powering modes are to being
The impact of system is shown in Table 2:
Table 2
Aggressive mode: fully functional pattern.The digital core of voltage-stablizer is opened, 16mhz rc oscillator or
32mhz crystal oscillator runs, or both runs；32khz rcosc oscillator or 32khz
Xosc oscillator runs.
Idle pulley: except the kernel of cc2541 chip (i.e. idle) out of service, other and aggressive mode
Equally.
The numerical portion of pm1: voltage-stablizer is opened, and 32mhz xosc and 16mhz rcosc does not run；
32khz rcosc or 32khz xosc runs；Reset, external interrupt or sleep timer expired when system
Will go to aggressive mode.
The digital core of pm2: voltage-stablizer is closed, and 32mhz xosc and 16mhz rcosc does not run；
32khz rcosc or 32khz xosc runs；Reset, external interrupt or sleep timer expired when system
Will go to aggressive mode.
The digital core of pm3: voltage-stablizer is closed, and all of oscillator does not run, when reset or external interrupt
System will go to aggressive mode.
According to system, each the feature of powering mode and the working condition under this application carry out low power operation
Model Design.Add low-power consumption sleeping task in the osal layer of program, for different running statuses pair
Service condition in system resource enters different low power operation patterns.Power consumption control method flow chart such as Fig. 3
Shown.
As preferred embodiment, described signal detection headend equipment 210 also includes and described main control device
215 connection accelerometers 217, can real time record wearer three directional accelerations, and pass through i2C bus
Issue main control chip cc2541, provide initial data for anti-tumble function.From mpu6050 as acceleration
Degree meter, its digital moving sensing incorporating three-axis gyroscope and three axis accelerometer producing for invensense
Processor.Compared to multicompartment scheme, invention avoids the between centers difference problem of combination gyroscope and accelerometer,
Decrease installing space, peripheral components are little.Accelerometer supply voltage is 2.5v, and operating current is only
500ua, accelerometer battery saving mode electric current 40ua@10hz, gyroscope operating current 5ma, gyroscope is treated
Dynamo-electric stream 5ua, is the qfn-24 encapsulation of 4mm × 4mm.Acceleration analysis scope can programme-control, have
± 2g, ± 4g, ± 8g and ± 16g.By i2C interface, melts to 6 complete axles of main control chip cc2541 output
Costar the evidence that counts.The motion process data bank being supplied to by invensense, can easily realize appearance
State resolves, and reduces the load to operating system for the motion process computing, greatly reduces development difficulty simultaneously.
Described mobile terminal device 220 includes middle control module 221 and connected signal characteristic abstraction module
222nd, physically different detection module 223 and emotion recognition module 224, described signal characteristic abstraction module 222
For extracting the signature waveform of described electro-physiological signals, described physically different detection module 223 is used for according to being carried
The signature waveform taking carries out physically different detection, and described emotion recognition module 224 is used for according to the feature extracted
Waveform carries out emotion recognition.
Described mobile terminal device 220 also include with the described middle control judge module 225 that is connected of module 221 and
Health consultation module 226, described judge module 225 is used for judging physically different testing result and emotion recognition knot
Really, described health consultation module 226 is used for when described judge module is judged physically different or produces the disturbance of emotion
When health consultation is provided and informs and carry out physically different detection again, otherwise inform that user's testing result is normal and moves back
Go out detection, the described disturbance of emotion is the Negative Affect such as sad, angry, detest and fear, that is, when described judgement
Module 225 judge physiological detection result be tachycardia, areocardia or VPB, or emotion recognition
When result is the Negative Affect such as sad, angry, detest, then described health consultation module 226 starts and is user
Offer an opinion consulting, rehabilitation suggestion and related service and remind user adopting health consultation suggestion after enter again
The physically different detection of row and emotion recognition, until the physiological detection result of user is normal and emotion recognition result is
Positive emotion；When described judge module 225 judges that physiological detection result and emotion recognition are normal, then institute
State health consultation module 226 and inform that this testing result of user is normal, detection can be exited.When implementing,
Described health consultation module 226 can be mated accordingly first according to the emotional state of user and physiology electric abnormal conditions
Issue-resolution, and and user carry out language interaction, further appreciate that other physiology of user and psychological feelings
Condition, is modified to original issue-resolution.The electro-physiological signals of synchronous detection user are simultaneously detected,
Judge whether the Physiological Psychology index of user is normal, and provide new solution further, until asking of user
The key to exercises is determined.
The present invention comprises physiology electric Outlier Detection Algorithm and Emotion identification algorithm.The former is initially with Moving Window method meter
Calculate the average r wave number in a period of time window, represent the heart rate of this time window end position with this numerical value, then
Physiology electric abnormal conditions are detected according to changes in heart rate situation.Can be to the holistic health shape of wearer in a period of time
Condition makees a corresponding assessment.The method that the latter adopts machine learning, the mood life having been marked in advance according to some
The electric sample data training linear classifier of reason, is mapped to higher-dimension physiology electrical characteristic data the most suitable by transition matrix
Spatially, the result after then being mapped according to sample is apart from all kinds of centers for the low level that can divide between inhomogeneity sample
The classification to judge mood signal for the distance of point, to realize carrying out the function of Emotion identification according to physiology electric, simultaneously
By to the perception of wearer's mood and the perceptible feedback of electro-physiological signals to mobile phone robot system, mobile phone robot
Interacted by second signals and user, and consulting is provided, adjusts and service.Present system power consumption
Extremely low, under 25 DEG C of environment, system operating voltage is 2.5v, system whole machine in the case of being not connected with main equipment
Running current is about 8ma, and when sending data, average current is about 12ma, and power consumption is significantly lower than like product.
Embodiment described above only have expressed the present invention preferred embodiment, and its description is more concrete and detailed
Carefully, but therefore can not be interpreted as the restriction to the scope of the claims of the present invention.It should be pointed out that for ability
For the those of ordinary skill in domain, without departing from the inventive concept of the premise, can also make some deformation and
Improve, these broadly fall into protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be with appended power
Profit requires to be defined.
Claims (10)
1. a kind of machine human feelings emotion identification method is it is characterised in that include step:
The electro-physiological signals of detection human body；
Extract the signature waveform of described electro-physiological signals；
Described signature waveform according to being extracted carries out physically different detection；
Using linear discriminent analysis lda algorithm, emotion recognition is carried out to the described signature waveform being extracted.
2. a kind of robot emotion identification method according to claim 1 is it is characterised in that described carry
Also include before the characteristic data step taking described electro-physiological signals:
Remove the unavailable data in described electro-physiological signals；
Electro-physiological signals described in normalized；
LPF is carried out to described electro-physiological signals.
3. a kind of robot emotion identification method according to claim 2 is it is characterised in that described carry
Take the characteristic of described electro-physiological signals method particularly includes:
Extract the pqrst signature waveform of human ecg signal.
4. a kind of robot emotion identification method according to claim 3 is it is characterised in that described
Carry out physically different detection according to the described characteristic extracted method particularly includes:
Using the inverse of phase between r popin is equal in certain time as heart rate；
Physically different detection is carried out according to described heart rate.
5. a kind of robot emotion identification method according to claim 3 or 4 is it is characterised in that institute
State, using lda algorithm, emotion recognition carried out to the described signature waveform being extracted method particularly includes:
Gather the signature waveform sample matrix x under variant emotion and set up emotional category vector v, described feelings
The vector that sense categorization vector is formed by variant emotion identification；
Pick out the sample matrix x of each classification according to described v respectively, calculate the sample of variant classification
Within class scatter matrix and total inter _ class relationship matrix:
..., k
Wherein, x is the sample data matrix of input, and behavior sample one by one is classified as feature one by one；It is all kinds of
Sample average；K is sample generic number；
Calculate scatter matrix between sample:
Wherein, it is Different categories of samples number, be sample overall mean；
Take the characteristic vector composition transition matrix w corresponding to front k-1 characteristic value of matrix；
By formula w*It is calculated the central point after each class projection；
New according to the far and near judgement of Euclidean distance of new feature samples all kinds of central point of distance after matrix w projection
The affiliated emotional category of sample data.
6. a kind of robot emotion recognition system is it is characterised in that include separate signal detection front end
Equipment and mobile terminal device, described signal detection headend equipment and mobile terminal device carry out radio communication even
Connect, described signal detection headend equipment is used for gathering and pre-processing Human Physiology electric signal and send to described shifting
Dynamic terminal device, described mobile terminal device is used for being entered according to the described pretreated electro-physiological signals being received
The physically different detection of row and emotion recognition.
7. a kind of robot emotion recognition system according to claim 6 is it is characterised in that described letter
Signal collecting device that number detection headend equipment includes being sequentially connected, multiplying arrangement, filter apparatus, ad conversion
Equipment and main control device, described signal collecting device is used for gathering electro-physiological signals, and described multiplying arrangement is used for
Amplify the electro-physiological signals of described collection, described filter apparatus are used for carrying out clutter to the electro-physiological signals after amplifying
Filter, described ad conversion equipment is used for filtered simulation electro-physiological signals are converted to data signal and are sent
To described main control device, described main control device receives described data signal and is wirelessly transmitted to described mobile whole
End equipment.
8. a kind of robot emotion recognition system according to claim 7 is it is characterised in that described signal
Detection headend equipment also includes the power supply management device being connected with described main control device, for described signal detection
Headend equipment carries out power supply power consumption management.
9. a kind of robot emotion recognition system according to claim 8 is it is characterised in that described movement
Terminal device include middle control module and connected signal characteristic abstraction module, physically different detection module with
And emotion recognition module, described signal characteristic abstraction module is used for extracting the signature waveform of described electro-physiological signals,
Described physically different detection module is used for carrying out physically different detection, described emotion according to the signature waveform being extracted
Identification module is used for carrying out emotion recognition according to the signature waveform being extracted.
10. a kind of robot emotion recognition system according to claim 9 is it is characterised in that described shifting
Dynamic terminal device also includes judge module and the health consultation module being connected with described middle control module, described judgement mould
Block is used for judging physically different testing result and emotion recognition result, and described health consultation module is used for described sentencing
Disconnected module is judged physically different or is produced and provide health consultation during the disturbance of emotion and inform that to carry out physiology again different
Often detection or emotion recognition, otherwise informs that user's testing result is normal and exits detection, the described disturbance of emotion is to disappear
Pole emotion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510198617.3A CN106338935A (en) | 2015-04-23 | 2015-04-23 | Robot emotion recognition method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510198617.3A CN106338935A (en) | 2015-04-23 | 2015-04-23 | Robot emotion recognition method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106338935A true CN106338935A (en) | 2017-01-18 |
Family
ID=57825829
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510198617.3A Pending CN106338935A (en) | 2015-04-23 | 2015-04-23 | Robot emotion recognition method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106338935A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113711149A (en) * | 2019-04-22 | 2021-11-26 | 松下知识产权经营株式会社 | Learned model generation method, maintenance display device, and program |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1771883A (en) * | 2004-11-11 | 2006-05-17 | 乐金电子(中国)研究开发中心有限公司 | Brain wave signal measuring system with mobile communication terminal |
| CN101930734A (en) * | 2010-07-29 | 2010-12-29 | 重庆大学 | Classification and identification method and device for cardiechema signals |
| CN102499676A (en) * | 2011-11-03 | 2012-06-20 | 北京工业大学 | Effective time sequence and electrode recombination based electroencephalograph signal categorizing system and method |
| CN204105953U (en) * | 2014-10-08 | 2015-01-21 | 鲍崇智 | A kind of pulse emotion recognition audio presentation system |
| CN104305964A (en) * | 2014-11-11 | 2015-01-28 | 东南大学 | Head mounted fatigue detector and method |
-
2015
- 2015-04-23 CN CN201510198617.3A patent/CN106338935A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1771883A (en) * | 2004-11-11 | 2006-05-17 | 乐金电子(中国)研究开发中心有限公司 | Brain wave signal measuring system with mobile communication terminal |
| CN101930734A (en) * | 2010-07-29 | 2010-12-29 | 重庆大学 | Classification and identification method and device for cardiechema signals |
| CN102499676A (en) * | 2011-11-03 | 2012-06-20 | 北京工业大学 | Effective time sequence and electrode recombination based electroencephalograph signal categorizing system and method |
| CN204105953U (en) * | 2014-10-08 | 2015-01-21 | 鲍崇智 | A kind of pulse emotion recognition audio presentation system |
| CN104305964A (en) * | 2014-11-11 | 2015-01-28 | 东南大学 | Head mounted fatigue detector and method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113711149A (en) * | 2019-04-22 | 2021-11-26 | 松下知识产权经营株式会社 | Learned model generation method, maintenance display device, and program |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105395192A (en) | Wearable emotion recognition method and system based on electroencephalogram | |
| CN103705214B (en) | A kind of healthy Mytip | |
| CN109480858A (en) | It is a kind of for quantify detect disturbances in patients with Parkinson disease bradykinesia symptom wearable intelligence system and method | |
| CN109009028B (en) | Wearable device capable of reflecting human fatigue degree | |
| CN105249943A (en) | An intelligent clothes control system and a control method thereof | |
| CN104720748A (en) | Sleep stage determining method and sleep stage determining system | |
| CN107361755A (en) | Intelligent watch with dysarteriotony prompting | |
| CN206007243U (en) | A kind of ECG detection device | |
| CN205514533U (en) | Wearable human skin detection device | |
| CN107669262A (en) | Multi-lead Telediagnosis of Electrocardiogram Signals and monitor system and method based on SVM and WLT | |
| CN106108878A (en) | A kind of intelligent watch for health monitoring | |
| CN108186024A (en) | A kind of movement gait monitor and multi-parameter sensor data processing method | |
| Scrugli et al. | An adaptive cognitive sensor node for ECG monitoring in the Internet of Medical Things | |
| CN102934998B (en) | Pressure evaluating system, relieving pressure control system and exercise pressure reduction control method | |
| CN103239218B (en) | Pulse monitoring equipment and system | |
| CN105982643A (en) | Sleep event detection method and system | |
| CN205041399U (en) | Intelligent electrocardiogram monitoring system | |
| CN104665837A (en) | Device for reminding wearer to strengthen neck exercise and control method thereof | |
| CN112263254A (en) | Human body energy consumption prediction system based on surface electromyogram signal sensor and prediction method thereof | |
| CN206414270U (en) | A kind of intelligent watch for health monitoring | |
| Wang et al. | Energy and accuracy trade-offs in accelerometry-based activity recognition | |
| CN206473320U (en) | EEG monitoring device and system | |
| CN106338935A (en) | Robot emotion recognition method and system | |
| CN209932728U (en) | Portable electrocardio monitoring clothing | |
| CN204192584U (en) | The intelligent individual soldier's motion monitoring of People's Armed Police and vital sign early warning salvage system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170118 |
|
| RJ01 | Rejection of invention patent application after publication |