CN110448281A - A kind of wearable work fatigue detection system based on multisensor - Google Patents

Abstract

The invention discloses a wearable work fatigue detection system based on multiple sensors, which includes a fatigue monitoring terminal device and a work fatigue analysis module. The fatigue monitoring terminal equipment uses the wearable sensor module to collect the original physiological signal, and sends it to the job fatigue analysis module through the ZigBee module; the data processing module preprocesses the original physiological signal and extracts the characteristic value, extracts the pulse rate in the pulse wave signal and calculates The blood pressure value extracts the single-step motion law in the acceleration signal, extracts the instantaneous body temperature value in the body temperature signal, and extracts the temperature, humidity and air pressure in the environmental parameters; the fatigue recognition module uses the obtained eigenvalues as the input of the GA-PNN neural network, and The output probabilities of the neural network are converted into corresponding fatigue levels. The invention realizes the real-time discrimination of the fatigue level, has the characteristics of high accuracy, strong anti-interference ability, etc., and can intuitively display the states and fluctuations of the concerned parameters without affecting the normal work of operators.

Description

A wearable work fatigue detection system based on multi-sensor

technical field

The invention relates to the field of human fatigue testing, in particular to a wearable work fatigue detection system based on multiple sensors.

Background technique

Fatigue has now become an important factor in accidents. An epidemiological survey in the United States found that among the American people, 14% of adult men and 29% of women have obvious symptoms of fatigue. In my country, the number of chronic diseases and even death caused by excessive fatigue caused by high work intensity and little rest time is also increasing year by year. Studies have found that in a large number of middle-aged intellectuals who have weakened their physiques and frequently suffered from chronic diseases, the causes are mostly "fatigue" caused by long hours of work, mental stress, lack of rest time, and physical exercise. According to expert fatigue surveys, a small half of the world's population is in a state of fatigue, accounting for more than 35% of the total population, and middle-aged men are as high as 60% to 75%. And among these crowds, the excessive fatigue of pilots and car drivers has caused most flight and traffic safety accidents. The method of assessing the fatigue state of workers based on physiological signal detection has the characteristics of objectivity and high detection accuracy, and is also the mainstream method of fatigue state detection.

Patent CN106530621A discloses a method and device for safe driving based on smart wearable devices. The system uses whether the heart rate and blood pressure signal exceeds the threshold to judge fatigue or not. This method has fewer monitoring parameters, single judgment conditions, and accuracy is difficult to guarantee .

Patent CN105662407A discloses a driver fatigue monitoring system based on surface myoelectric technology, which uses detection electrodes to monitor surface myoelectric signals within the eyelid range to obtain eye opening and closing states. This method requires electrodes to be installed, cannot be used for a long time, and will affect normal driving.

Contents of the invention

The purpose of the present invention is to provide a wearable work fatigue detection system based on multi-sensors, which can be fused according to multi-sensor data to judge the fatigue degree of the current operator, and send different early warning signals according to the current fatigue state, with high fault tolerance rate, It has good real-time performance and does not affect normal operation.

The technical solution to realize the present invention is: a wearable work fatigue detection system based on multi-sensors, including

Fatigue monitoring terminal equipment used to collect data parameters and send the data parameters;

And the fatigue analysis module is used to receive the collected data parameters and perform parameter processing, and establish a fatigue level recognition model on the basis of various eigenvalue data of the processed training samples, and use various characteristics of the operator The parameters are used as the input of the trained recognition model to obtain the operator's fatigue level in real time.

The fatigue monitoring terminal equipment includes

The main control module controls the data collection of each sensor detection module and the data transmission with the work fatigue analysis module. By configuring the sensor register, the sensor is initialized; at the same time, the internal register data of the sensor is read to identify the sensor status information and signal data.

The pulse and blood pressure module connected to the main control module is placed at the bottom of the wristband and directly contacts with the skin. It uses photoelectric plethysmography to sense and extract the pulse information of the human body, and outputs the pulse wave signal.

The body temperature module connected to the main control module is placed at the bottom of the wristband and directly contacts the skin to obtain the skin temperature of the operator's arm.

The acceleration module that communicates with the main control module through the I2C digital interface is placed inside the wristband to detect the three-axis acceleration signal generated by the hand swing and record the changes in the entire movement process.

The environmental parameter module connected with the main control module is placed inside the wristband to monitor three signals of ambient temperature, humidity and atmospheric pressure.

Compared with the prior art, the present invention has significant advantages in that:

(1) Less affected by single sensor noise and error, low misjudgment rate and high stability. (2) The wristband-style wearable design can move with the operator without affecting normal work and life. (3) The fatigue level can be given in real time, with good immediacy.

Description of drawings

Fig. 1 is a frame diagram of the multi-sensor-based wearable fatigue detection system of the present invention.

Fig. 2 is a data processing flow chart of the work fatigue analysis module of the present invention.

Fig. 3 is a flowchart of training samples, paired samples and detection in the present invention.

Detailed ways

Referring to FIG. 1 , the present invention provides a multi-sensor-based wearable job fatigue detection system, including a fatigue monitoring terminal device and a job fatigue analysis module. Fatigue monitoring terminal equipment includes pulse blood pressure module 1, body temperature module 2, acceleration module 3, environmental parameter module 4 and main control module 5, main control module 5 is connected with pulse blood pressure module 1, body temperature module 2, acceleration module 3, environmental parameter module respectively 4 connection; the job fatigue analysis module includes a ZigBee module, a data processing module, and a fatigue recognition module connected in sequence.

The fatigue monitoring terminal equipment is integrated on the bracelet.

The pulse and blood pressure module 1 is placed at the bottom of the wristband, in direct contact with the skin, and uses photoelectric plethysmography (PPG) to sense and extract the pulse information of the human body, output the pulse wave signal, and send it to the main control module 5 .

The body temperature module 2 is placed at the bottom of the bracelet and directly contacts the skin to obtain the skin temperature of the operator's arm. Both the pulse and blood pressure module 1 and the body temperature module 2 are flush with the bottom surface of the bracelet, and there will be no discomfort during the entire use process. .

The acceleration module 3 is placed inside the wristband, which mainly detects the three-axis acceleration signal generated by the hand swing, records the changes in the entire movement process, and communicates with the main control module 5 through the I2C digital interface.

The environmental parameter module 4 is placed inside the wristband, which mainly monitors three signals of ambient temperature, humidity and atmospheric pressure. The wristband will reserve a small hole to connect to the outside world to prevent the airtight space from affecting the measurement of environmental parameters.

The main control module uses the CC2530 chip as the main control chip to control the data collection of each sensor detection module and the data transmission with the job fatigue analysis module. Identify sensor status information and signal data. The chip has the characteristics of high performance, low cost, and low power consumption, which can fully meet the system requirements.

The data transmission with the work fatigue analysis module uses ZigBee to realize wireless communication, and the ceramic antenna is used to reduce the size of the wristband. The data acquisition program of the fatigue monitoring terminal equipment and the supporting ZigBee communication protocol are all carried on the CC2530 chip, which reduces the use of productization. The complexity of the circuit design of the wireless communication module.

The work fatigue analysis module is installed on the PC side, so that the work fatigue analysis module communicates with the main control module through the control serial port, and sends and receives sensor data and control instructions.

Combined with Figure 2, the job fatigue analysis module sends instructions to select the type of monitoring data, and the fatigue monitoring terminal equipment collects data according to the instructions of the job fatigue analysis module, and packages and sends them. Through the received raw human physiological data, the data processing module of the job fatigue analysis module segments the raw human physiological data, and obtains the characteristic signals of various parameters in each small period of time:

(1) Use the EMD soft threshold method to remove the interference of noise on the pulse wave signal, select the pulse rate and amplitude as the characteristic value of the pulse wave signal, set the threshold and pair it according to the positive and negative values of the derivative at the threshold, a complete pulse signal Contains the positive and negative derivative values once each to obtain the pulse rate;

(2) Remove the interference of noise on the pulse wave signal by EMD soft threshold method, extract the pulse wave pulse width, and calculate the systolic blood pressure based on the relationship between the pulse wave transmission velocity, blood vessel elastic modulus and blood vessel wall pressure; extract the pulse Based on the peak and valley values of the wave signal, the diastolic pressure is calculated based on the pulse characteristic K value theory and the elastic cavity model;

(3) Filter out the noise interference through the second-order Chebyshev filter, and solve the average value of the temperature signal as a feature in a short period of time;

(4) Use wavelet analysis to denoise the acceleration signal, and use the autocorrelation coefficient of the acceleration signal as the feature of the arm movement but the single-step law;

(5) Use the second-order Chebyshev filter to filter out the interference of noise, and solve the average value of the temperature, humidity and air pressure signals as the characteristics in a short period of time.

Each group of signals includes pulse, blood pressure, body temperature, acceleration, and environment five types of eigenvalues in the same short period of time. The work fatigue analysis module realizes the comprehensive display of real-time data and eigenvalues of various sensors.

The fatigue recognition module establishes a fatigue level recognition model with high reliability and high fault tolerance on the basis of various eigenvalue data of the processed training samples. Input to obtain the fatigue level of the operator in real time.

Combined with Figure 3, before the formal discrimination, the fatigue monitoring terminal equipment is used to collect samples of pulse, skin temperature signal, motion acceleration and environmental parameter signals under fatigue state and normal state, based on the objective evaluation index of fatigue state and the description of fatigue qualitative characteristics, the establishment of A sample database of fatigue state evaluation benchmarks.

Step 1-1: Select multiple observed subjects, collect physiological parameters and perform PVT tests on them, and obtain raw data of physiological parameters and quantitative fatigue levels. The collection of physiological parameters uses the fatigue monitoring terminal equipment shown in Figure 1, wears the wristband on the wrist, and collects physiological parameters for 5 minutes. The eigenvalues of all data after preprocessing are used as the input of training samples; based on individual reaction time Psychomotor and focused attention, the psychomotor vigilance task (Psychomotor Vigilance Task, PVT) is used to quantify the individual's fatigue level, which is one of the most commonly used fatigue measurement methods in sleep deprivation and behavioral performance research. During the experiment, when the target was stimulated (for example, when a red ball appeared in the center of a black computer screen), the observed person reacted immediately, quickly pressing the button to respond to the visual stimulation, early operation, wrong button, and button duration were all considered. For wrong operation, when the operator is tired, it will lead to an increase in reaction time, a decrease in reaction speed, and an increase in error rate. The test indicators include the reaction time of the target visual stimulus, the number of missed times, and the number of various errors. Passed These indicators reflect the level of fatigue. Each observed subject performs a 5-minute PVT test to obtain the corresponding fatigue level, which is used as the output of the training sample.

Step 1-2: pair the physiological parameters obtained from the experiment with the fatigue level obtained from the PVT test to form a training sample, and establish a training sample database. The z-score data standardization method is used to normalize data of different orders of magnitude to eliminate the dimensional influence between indicators.

The GA-PNN neural network is trained by using the input and output of the processed training samples. The model expression of the PNN neural network is net=newpnn(P,T,spread), where P represents the input sample vector, which corresponds to the input data sample composed of multi-sensor parameters, and T is the output sample vector, which corresponds to the fatigue quantization level. The spread value of the network expansion constant needs to pass different tests to obtain the optimal spread value. The spread value is optimized using the genetic algorithm, and an evolutionary algebraic counter with a number of 200 is established in the genetic (GA) algorithm, and 20 individuals are randomly generated as the initial population P(0), and the output error of the PNN neural network is used as the fitness function. Initialize the fitness value of the population, use selection, crossover, and mutation operations to obtain optimized individuals with high fitness or pass on new individuals generated by paired crossover to the next generation, and calculate and evaluate the individual fitness after each generation of evolution. When the evolution process is combined The obtained maximum fitness individual is used as the optimal output, and this optimal value is the optimal spread value of the PNN neural network; finally, after training, the threshold value of each node in the hidden layer and the connection weight of the input layer, as well as the hidden layer and the output layer are obtained. Connection weights between layers to get a trained neural network.

Take the training sample data as the test sample, bring the input of the training sample into the trained neural network, compare the output result obtained with the original fatigue quantification level, if the output result is consistent with the expected result, it means that the network has The training is complete, otherwise modify the parameters and return to step 3 for retraining.

A detection method based on a multi-sensor wearable work fatigue detection system, specifically comprising the following steps:

Step 1: When it is officially used, use the fatigue monitoring terminal equipment to collect pulse, skin temperature signal, motion acceleration and environmental parameter signal again, and pack various data and transmit them to the job fatigue analysis module through the ZigBee module;

Step 2: The job fatigue analysis module standardizes the z-score data on the received data, uses the obtained test sample data as the eigenvalue input vector of the neural network, and finally obtains the fatigue probability output value, and judges the current status of the operator according to the fatigue probability. Fatigue level, according to the characteristics of the PNN neural network, the larger the number of samples, the smaller the error in the analysis and identification of job fatigue status.

Step 3: The job fatigue analysis software gives corresponding reminders and early warnings to the operators according to the obtained fatigue level.

Claims (7)

1. a kind of wearable work fatigue detection system based on multisensor, it is characterised in that: including

For acquiring the fatigue monitoring terminal device of data parameters, and data parameters are sent；

With analysis of fatigue module, data parameters for collecting, and parameter processing is carried out, and in the instruction handled well On the basis of all kinds of characteristic value datas for practicing sample, level of fatigue identification model is established, with all kinds of characteristic parameters of operating personnel As the input of trained identification model, the level of fatigue operation of operating personnel is obtained in real time.

2. the wearable work fatigue detection system according to claim 1 based on multisensor, it is characterised in that: described Fatigue monitoring terminal device includes

Main control module controls the data acquisition of each sensor detection module and transmits with the data of work fatigue analysis module, By way of sensors configured register, initializing sensor work；Sensor internal register data is read simultaneously, to know Individual sensor status information and signal data,

The pulse blood pressure module connecting with main control module is placed in bracelet bottom, in direct contact with the skin, using photo-electric volume arteries and veins The mode that wave of fighting is traced incudes the pulse information of human body and is extracted, and exports pulse wave signal,

The body temperature module connecting with main control module is placed in bracelet bottom, in direct contact with the skin, obtains the table of operating personnel's arm Skin temperature,

The acceleration module communicated by I2C digital interface with main control module, is placed in inside bracelet, and detection hand swings generation 3-axis acceleration signal records the variation of entire motion process,

The environmental parameter module connecting with main control module, is placed in inside bracelet, monitoring environment temperature, humidity and three kinds of atmospheric pressure Signal.

3. the wearable work fatigue detection system according to claim 1 based on multisensor, it is characterised in that: described Fatigue monitoring terminal device is integrated on bracelet.

4. the wearable work fatigue detection system according to claim 2 based on multisensor, it is characterised in that: master control Module uses CC2530 chip, and the data transmission with work fatigue analysis module is realized using ZigBee to be wirelessly communicated, and data are adopted Collection program and matched ZigBee communication agreement are mounted in CC2530 chip.

5. the wearable work fatigue detection system according to claim 1 based on multisensor, it is characterised in that: fatigue Analysis module is equipped on the end PC, including sequentially connected ZigBee module, data processing module, tired identification module.

6. the wearable work fatigue detection system according to claim 5 based on multisensor, it is characterised in that:

Data processing module sends instruction selection monitoring data type by ZigBee module, and fatigue monitoring terminal device is according to finger Acquisition data are enabled, are transmitted, by the Human Physiology initial data received, data processing module is to Human Physiology original number According to being segmented, and obtain the characteristic signal of all kinds of parameters in each short time:

1) interference of the signal to pulse wave signal is removed by EMD Soft thresholding, selects pulse frequency and amplitude as pulse wave signal Characteristic value, setting threshold value simultaneously matched according to the derivative positive and negative values at threshold value, and primary complete pulse signal includes positive and negative Derivative value is each primary, obtains pulse frequency with this；

2) interference of the signal to pulse wave signal is removed by EMD Soft thresholding, extracts pulse wave pulsewidth, transmitted based on pulse wave Relationship between speed, blood vessel elasticity modulus and vascular wall pressure three calculates systolic pressure；Extract the wave crest of pulse wave signal And valley value, diastolic pressure is calculated based on pulse characteristics K value theory and elastic chamber model；

3) interference that noise is filtered out by second order Chebyshev filter solves the average value of temperature signal as in short time Feature；

4) it selects wavelet analysis to carry out denoising to acceleration signal, acceleration signal auto-correlation coefficient is used to transport as arm Dynamic but single step law characteristic；

5) interference of noise is filtered out by second order Chebyshev filter, the average value for solving temperature, humidity and air pressure signal is made For the feature in short time；

Every group of signal includes pulse, blood pressure, body temperature, acceleration, five category feature value of environment in same short time, is realized various The synthesis display of sensor real time data and characteristic value.

7. the wearable work fatigue detection system according to claim 5 based on multisensor, it is characterised in that: fatigue It is high to establish high reliablity, serious forgiveness on the basis of all kinds of characteristic value datas for the training sample handled well for identification module Level of fatigue identification model, it is real using all kinds of characteristic parameters of operating personnel as the input of trained identification model When obtain operating personnel level of fatigue.