Wearable smart monitoring method and device for emotions of construction industry workers

Abstract

The application discloses a kind of wearable assembled building industry worker emotional intelligent monitoring method and device;Its technical points are in, S100, physiological signal is collected;The physiological signal includes: electrocardiogram signal;S200, the electrocardiogram signal obtained by S100 is carried out denoising processing;S300, extract heart rate variability HRV characteristics from electrocardiogram signal, including time domain feature and frequency domain feature;S400, the extracted HRV characteristics are passed through trained SVM-KNN classifier, to identify emotional category;S500, the identification result is output to handheld end.Using a kind of wearable assembled building industry worker emotional intelligent monitoring method and device, the emotion of industry worker can be effectively identified.

Description

Technical Field

[0001] This invention relates to the field of wearable equipment technology, and more specifically, to a wearable method and device for intelligent monitoring of the emotions of workers in the prefabricated building industry. Background Technology

[0002] Regarding emotion monitoring technology, the following existing technologies exist:

[0003] [1] China Mobile Communications Group Corporation. A method and system for monitoring emotions - 201410637181.9

[0004] [2] China Mobile Communications Group Shandong Co., Ltd. A method and system for determining intelligent emotions - 201510613689.X.

[0005] [3] Shanghai Broadband Technology and Application Engineering Research Center, Shanghai Xiaotong Tiandi Information Technology Co., Ltd. A method for judging emotions based on heart rate and respiration - 201510075507.8.

[0006] The aforementioned existing technologies all rely on various physiological signals such as heart rate, respiration, and audio to determine emotions. First, physiological signals are monitored on the participants for a continuous period. During the experiment, different actions are performed to induce different emotional states. The participants' different emotional states and the physiological signal parameters under those emotional states are recorded, and an evaluation report is generated. Then, physiological signal monitoring is performed on the subjects being evaluated, and the obtained data is input into the emotion evaluation report to determine the subjects' emotional states, thus achieving real-time monitoring of human emotions.

[0007] Among the factors influencing unsafe behaviors in prefabricated building construction, emotion is a significant one. Existing research on affective computing largely relies on specialized, medical-grade signal acquisition devices. While these devices can accurately capture physiological signals, they often fail to meet wearable requirements. Furthermore, most wearable devices, although collecting physiological signals, only perform simple statistical analysis and cannot provide accurate emotion recognition. Therefore, current emotion recognition technologies for prefabricated building workers cannot meet the needs of practical engineering projects. Summary of the Invention

[0008] The purpose of this invention is to address the shortcomings of the prior art by providing a wearable method and device for intelligent monitoring of the emotions of workers in the prefabricated building industry.

[0009] A wearable method for intelligently monitoring the emotions of workers in the prefabricated building industry, comprising the following steps:

[0010] S100, Acquire physiological signals; the physiological signals include: electrocardiogram (ECG) signals;

[0011] S200 performs noise reduction processing on the electrocardiogram signal obtained from S100;

[0012] S300 extracts heart rate variability (HRV) features from electrocardiogram (ECG) signals, including time-domain and frequency-domain features.

[0013] S400 extracts HRV features and passes them through a trained SVM-KNN classifier to identify the emotion category.

[0014] The S500 outputs the recognition results to the handheld device.

[0015] Furthermore, the S300 employs the pan_tompkin algorithm to detect QRS complex information and extract R-wave peak information; it then extracts heart rate variability (HRV) features from this information and statistically analyzes the changes between R waves.

[0016] HRV(i) = R(i+1) - R(i), where i is the sampling point corresponding to the peak position of R wave;

[0017] Sixteen HRV feature parameters were extracted from the RR interval (i.e., the interval between each heartbeat), including nine time-domain features and seven frequency-domain features, thus forming the feature set for emotion recognition;

[0018] (1) The time-domain characteristics of HRV are as follows:

[0019] SDNN: Standard deviation of all RR intervals;

[0020] RMSSD: Root mean square of the difference between adjacent RR periods;

[0021] SDNNindex: The average of the standard deviations of all RR intervals;

[0022] SDSD: Standard deviation of the difference between adjacent RR values;

[0023] NN50 count: The number of RR intervals that differ by more than 50ms;

[0024] PNN50: The number of RR50 divided by the sum of all RR intervals;

[0025] Hrviendex: Total number of RR intervals divided by the height of the RR interval;

[0026] Max-min: The difference between the maximum and minimum values ​​among all RR intervals;

[0027] MEAN: The average RR interval;

[0028] (2) The HRV frequency domain characteristics are as follows:

[0029] totalpower: total power spectral density;

[0030] HF: High-frequency power, range 0.15~0.4Hz;

[0031] LF: Low-frequency power, range 0.04~0.15Hz;

[0032] VLF: Ultra-low frequency power, range 0.04~0.15Hz;

[0033] LF / HF: The ratio of low-frequency power to high-frequency power;

[0034] HFn: Normalization of high-frequency capability;

[0035] LFn: The proportion of low-frequency power in the total power spectrum.

[0036] Furthermore, the S400 also includes: passing the extracted ECG HRV features through a pre-trained SVM-KNN classifier to obtain the wearer's corresponding emotion type;

[0037] SVM uses kernel functions to map input data into a high-dimensional space, performs linear classification in the high-dimensional space, and thus constructs the optimal hyperplane for classifying the data.

[0038] KNN groups the test sample and the K nearest-weight classes in the training samples into one class;

[0039] (1) SVM classification function:

[0040] Wherein, the training set T = {(x1,y1),(x2,y2),...,(x N ,y N )};y i ∈Y={-1,1},i=1,2,...,N;

[0041] α i is the Lagrange multiplier; b is the threshold; C is the penalty factor;

[0042] γ is the kernel function parameter;

[0043] (2) KNN classification: All samples correspond to points in m-dimensional space, x i Let x represent the value of the i-th feature of x. i and x j The Euclidean distance is Select k sample points with the smallest distance values ​​to form a sample set, and count the categories of the samples in this sample set. The category with the most samples is the category of the sample to be tested.

[0044] (3) The SVM-KNN classification method is:

[0045] Calculate the test sample x and the representative points x of the two support vectors. + and x - The distance difference g(x) = ∑yiα i k(x,xi)+b;

[0046] The absolute value of g(x) is compared with the threshold. If it is greater than the threshold, the SVM classification result is output directly; if the absolute value is less than the threshold, the KNN classification algorithm is used to obtain the result.

[0047] A wearable intelligent monitoring device for the emotions of prefabricated building industry workers, employing the aforementioned monitoring method.

[0048] The advantages of this application are:

[0049] First, the basic concept of this application is: adopting the basic concept of "ECG signal - HRV features - SVM - KNN classifier - emotion recognition". This application proposes for the first time an "HRV feature set", including SDNN, RMSSD, SDNNindex, SDSD, NN50count, PNN50, Hrviendex, Max-min, MEAN; totalpower, HF, LF, VLF, LF / HF, HFn, and LFn. That is, it is feasible to use the above feature set for emotion recognition.

[0050] Second, the SVM classifier can accurately classify samples that are far from the hyperplane, while the KNN algorithm performs well in classifying sample data with overlapping class domains. Combining the SVM and KNN algorithms can improve the accuracy of emotion classification. Attached Figure Description

[0051] The present invention will be further described in detail below with reference to the embodiments shown in the accompanying drawings, but this does not constitute any limitation on the present invention.

[0052] Figure 1 This is a flowchart of a wearable prefabricated building industry worker emotion intelligence monitoring method. Detailed Implementation

[0053] Example 1: A novel wearable intelligent emotion monitoring technology and device for prefabricated building workers.

[0054] Unsafe behaviors of construction workers are a major cause of construction safety accidents, and the management of construction workers' safe behaviors is also a core issue in current safety production management. Negative emotions of construction workers may induce unsafe behaviors.

[0055] Example 1 proposes using electrocardiogram signals to monitor the emotional changes of construction workers in real time, improves the emotion recognition algorithm, and can also provide suggestions for construction workers to regulate their own emotions.

[0056] A novel wearable method for intelligently monitoring the emotions of workers in the prefabricated building industry includes the following steps:

[0057] Step 1: Collect electrocardiogram (ECG) signals using a wearable wristband without disrupting the wearer's daily life;

[0058] The second step is to transmit the collected ECG signals to a remote server in real time for preprocessing, designing filters to select the original ECG signals, and then denoising the ECG signals.

[0059] (1) Wavelet decomposition of the signal: Select a wavelet and determine the level N of wavelet decomposition according to the Nyquist sampling theorem, and then perform N-level wavelet decomposition on the signal.

[0060] (2) Detail processing: threshold quantization of high-frequency coefficients in wavelet decomposition.

[0061] In the wavelet transform domain, find all peak points and put them into the sigmax matrix; sort the data in the sigmax matrix, and use 25% to 55% of the difference between the mean of the first 8 maxima and the last 50 minima as the threshold.

[0062] There are two methods for threshold processing:

[0063] Soft thresholding—compares the absolute value of the signal with a threshold, sets points less than or equal to the threshold to zero, and changes the value of points greater than the threshold to the difference between the value of that point and the threshold.

[0064] Hard thresholding—compares the absolute value of the signal with a threshold, sets points less than or equal to the threshold to zero, and leaves points greater than the threshold unchanged.

[0065] (3) Wavelet reconstruction. Based on the low-frequency coefficients Can of the Nth layer of wavelet decomposition and the high-frequency coefficients from the 1st to the Nth layer after quantization, wavelet reconstruction of the signal is performed, thereby extracting the denoised ECG signal.

[0066] Step 3: The pan_tompkin (PT) algorithm is used to detect QRS complex information and extract R-wave peak information. Heart rate variability (HRV) features are then extracted from this information, and the changes between R waves are statistically analyzed: HRV(i) = R(i+1) - R(i), where i is the sampling point corresponding to the R-wave peak position. Sixteen HRV feature parameters are extracted through the RR interval, including nine time-domain features and seven frequency-domain features, thus forming the feature set for emotion recognition.

[0067] (1) The time-domain characteristics of HRV are as follows:

[0068] SDNN: Standard deviation of all RR intervals;

[0069] RMSSD: Root mean square of the difference between adjacent RR periods;

[0070] SDNNindex: The average of the standard deviations of all RR intervals;

[0071] SDSD: Standard deviation of the difference between adjacent RR values;

[0072] NN50 count: The number of RR intervals that differ by more than 50ms;

[0073] PNN50: The number of RR50 divided by the sum of all RR intervals;

[0074] Hrviendex: Total number of RR intervals divided by the height of the RR interval;

[0075] Max-min: The difference between the maximum and minimum values ​​among all RR intervals;

[0076] MEAN: The average RR interval;

[0077] (2) The HRV frequency domain characteristics are as follows:

[0078] totalpower: total power spectral density;

[0079] HF: High-frequency power, range 0.15~0.4Hz;

[0080] LF: Low-frequency power, range 0.04~0.15Hz;

[0081] VLF: Ultra-low frequency power, range 0.04~0.15Hz;

[0082] LF / HF: The ratio of low-frequency power to high-frequency power;

[0083] HFn: Normalization of high-frequency capability;

[0084] LFn: The proportion of low-frequency power in the total power spectrum;

[0085] Step 4: The extracted HRV (Human Respiratory Rate) features are processed by a pre-trained SVM-KNN classifier to obtain the wearer's corresponding emotion type. SVM uses a kernel function to map the input data into a high-dimensional space, performing linear classification within that space to construct the optimal hyperplane for data classification. KNN groups the test sample with the K nearest-weight classes from the training samples.

[0086] (1) SVM classification function:

[0087] Wherein, the training set T = {(x1,y1),(x2,y2),...,(x N ,y N )};y i ∈Y={-1,1},i=1,2,...,N;

[0088] α i is the Lagrange multiplier; b is the threshold; C is the penalty factor;

[0089] γ is the kernel function parameter.

[0090] (2) KNN classification principle: All samples correspond to points in m-dimensional space, x i Let x represent the value of the i-th feature of x. i and x j The Euclidean distance is Select k sample points with the smallest distance values ​​to form a sample set, and count the categories of the samples in this sample set. The category with the most samples is the category of the sample to be tested.

[0091] (3) SVM-KNN classification algorithm process: Calculate the test sample x and the representative points x of the two support vectors. + and x - The distance difference g(x) = ∑y i α i k(x,x i )+b. Compare the absolute value of g(x) with a threshold (usually set to 0.4~0.8). If it is larger than the threshold, the SVM classification result is directly output; if the absolute value is smaller than the threshold, the KNN classification algorithm is used to obtain the result.

[0092] Step 5: The identified results are output to the mobile app, and suggestions are provided for the wearer to regulate their emotions.

[0093] The above-described embodiments are preferred embodiments of the present invention and are only used to facilitate the illustration of the present invention. They are not intended to limit the present invention in any way. Any person skilled in the art who makes local modifications or alterations to the technical content disclosed in the present invention without departing from the scope of the technical features of the present invention shall still fall within the scope of the technical features of the present invention.

Claims

1. A wearable intelligent monitoring method for the emotions of workers in the prefabricated building industry, characterized in that, It includes the following steps: S100, for acquiring electrocardiogram (ECG) signals; S200, for denoising ECG signals; S300 extracts heart rate variability (HRV) features from electrocardiogram (ECG) signals, including time-domain and frequency-domain features. The pan_tompkin algorithm is used to detect QRS complexes and extract R wave peaks. Heart rate variability (HRV) features are then extracted from these R waves, and the changes between R waves are statistically analyzed: HRV(i) = R(i+1) - R(i), where i is the sampling point corresponding to the R wave peak position. Sixteen HRV feature parameters were extracted using the RR interval to construct a feature set for emotion recognition; the RR interval is the time interval between each heartbeat. The temporal features include: the standard deviation of all RR intervals SDNN, the root mean square of the difference between adjacent RR intervals RMSSD, the average standard deviation of all RR intervals SDNNindex, the standard deviation of the difference between adjacent RR intervals SDSD, the number of RR intervals with a difference of more than 50ms NN50 count, the number of RR50 intervals divided by the sum of all RR intervals PNN50, the total number of RR intervals divided by the height of the RR intervals Hrviendex, the difference between the maximum and minimum values ​​of all RR intervals Max-min, and the average value of the RR intervals MEAN; Frequency domain characteristics include: total power spectral density (totalpower), high-frequency power (HF) from 0.15 to 0.4 Hz, low-frequency power (LF) from 0.04 to 0.15 Hz, ultra-low-frequency power (VLF) from 0.04 to 0.15 Hz, the ratio of low-frequency to high-frequency power (LF / HF), normalized high-frequency capability (HFn), and the proportion of low-frequency power in the total power spectrum (LFn). S400 processes the HRV features through a trained SVM-KNN classifier to obtain the wearer's corresponding emotion category. The S500 outputs the recognition results to the handheld device. S400 also includes: SVM maps the input data into a high-dimensional space through a kernel function, performs linear classification in the high-dimensional space, and thus constructs the optimal hyperplane for classifying the data; KNN groups the test sample and the K nearest-weight classes in the training samples into one class; (1) SVM classification function: ; Wherein, the training set T={(x1,y1),(x2,y2),...,(x N ,y N )};y i ∈Y={-1,1},i=1,2,...,N; α i is the Lagrange multiplier; b is the threshold; C is the penalty factor; γ is the kernel function parameter; (2) KNN classification: All samples correspond to points in m-dimensional space, x i Let x represent the value of the i-th feature of x. i and x j The Euclidean distance is Select k sample points with the smallest distance values ​​to form a sample set, and count the categories of the samples in this sample set. The category with the most samples is the category of the sample to be tested. (3) The SVM-KNN classification method is: Calculate the test sample x and the representative points x of the two support vectors. + and x - Distance difference ; The absolute value of g(x) is compared with the threshold. If it is greater than the threshold, the SVM classification result is output directly; if the absolute value is less than the threshold, the KNN classification algorithm is used to obtain the result.

2. A wearable intelligent monitoring device for the emotions of prefabricated building industry workers, employing the monitoring method described in claim 1.

Images