CN111597877A - A fall detection method based on wireless signal - Google Patents

Abstract

A fall detection method based on wireless signals, which uses ordinary commercial WiFi equipment to establish a fall detection environment. When the target object is active in the detection area covered by WiFi, it will affect the wireless signal. By obtaining the channel state information in the received signal , using the deep learning network to process the CSI data to identify whether the target object falls in the detection area. The present invention can achieve high fall detection accuracy while satisfying convenience, ease of use and safety, and does not require the target to carry any special equipment, does not record user privacy, and is convenient and easy to deploy.

Description

A fall detection method based on wireless signal

technical field

The invention relates to the technical field of intelligent monitoring, in particular to a fall detection method based on a wireless signal.

Background technique

Data show that since my country entered an aging society at the end of the 20th century, the number of elderly people and their proportion in the total population has continued to grow. From 2000 to 2018, the number of elderly people aged 60 and above increased from 126 million to 249 million, and the elderly population accounted for 249 million. The share of the total population rose from 10.2% to 17.9%. In the future, the degree of aging will continue to deepen. According to the analysis of relevant data, falls are one of the important factors leading to disability and even death of the elderly. Timely detection of falls of the elderly can not only reduce medical expenses, but also enable the elderly to receive timely assistance.

At present, the commonly used method for human fall detection at home and abroad is to determine whether the target object falls through wearable sensor equipment or video image equipment. However, the wearable sensor-based method requires the target object to carry the sensor with them, which may cause inconvenience in the daily life of the elderly, and the cost of this method is relatively high. However, the methods based on video images are affected by factors such as lighting conditions and object occlusion, and may also leak personal privacy.

SUMMARY OF THE INVENTION

In view of the problems existing in the prior art, the present invention provides a fall detection method based on a wireless signal, which utilizes ordinary commercial WiFi equipment to detect whether a human body falls by analyzing different change patterns of channel state information, which can achieve higher recognition Accuracy, convenience and safety are met at the same time.

For achieving the above-mentioned goals, the technical scheme adopted in the present invention is:

A fall detection method based on a wireless signal, comprising the following steps:

Step 1: collect the channel state information in the wireless signal, and obtain the original CSI data;

Step 2: preprocessing the CSI original data to obtain a data set to be identified;

Step 3: Divide the CSI data set to be identified into two parts, one part is used as a training set and the other part is used as a test set;

Step 4: Import the training set in Step 3 into the deep neural network for feature extraction training, and use the test set to perform a fall recognition test on the trained neural network;

Step 5: Collect the CSI raw data of the wireless signal in the detection area in real time, and perform the preprocessing of step 2, import the data set to be identified into the neural network in step 4 for processing, and determine whether the target falls in the detection area.

Further, in the step 1, the CSI original data includes CSI amplitude data.

In the step 2, the CSI raw data is preprocessed to obtain the data set to be identified, including denoising the CSI raw data by using the singular spectrum analysis SSA algorithm, and denoising the denoised data through the Hilbert transform HHT. The original CSI data of the CSI is converted into a spectrogram, and the fallen or pseudo-fall CSI data is extracted from the spectrogram as the data set to be identified.

In the step 4, the deep neural network includes a convolutional neural network CNN.

Further, the structure of the deep neural network includes a generation module, a feature learning module and an output module:

The generation module is used to convert the CSI data into a feature map with a spatial coding mode;

A feature learning module is used to map spatially encoded patterns to features that recognize falls;

The output module is used to output corresponding recognition results according to the extracted features.

The convolutional neural network CNN includes eight deconvolution layers, a backbone network, a fully connected layer, an activation function and a loss function;

The backbone network includes ResNet, Inception and VGG network, the activation function includes SoftMax activation function, and the loss function includes cross entropy function.

Further, the SoftMax activation function is:

Among them, i represents the ith sample in the data set, K represents the number of samples in the data set, and represents the probability value of the ith sample, and the sum of these probability values is 1.

The cross-entropy loss function is:

Among them, N represents the number of samples i, represents the label of sample i (positive class is 1, negative class is 0), and represents the probability that sample i is predicted to be a positive class.

The beneficial effects of the present invention are:

1. Use ordinary commercial WiFi equipment to establish a detection environment, obtain the CSI value of the wireless signal according to the changes of the WiFi signal by the different actions of the target object, and analyze and process to detect whether the target falls;

2. The present invention can achieve high recognition accuracy while satisfying convenience, ease of use and safety, and does not require the human body to carry any special equipment, does not need to purchase other sensor equipment, etc., is convenient and easy to deploy, and has high security. specialty;

3. The present invention combines wireless signal and deep learning technology to apply to fall detection, which provides a new research idea for the field of medical monitoring.

Description of drawings

Figure 1 is an overall block diagram of fall detection;

Fig. 2 is the deep learning network frame diagram in the present invention;

FIG. 3 is a schematic diagram of the detection environment.

Detailed ways

The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, and the protection scope of the present invention can be more clearly defined.

1 to 3 , a method for detecting a fall based on a wireless signal includes the following steps:

Step 1: collect channel state information in the wireless signal, and obtain CSI original data including CSI amplitude data;

Step 2: preprocessing the CSI original data to obtain a data set to be identified;

The preprocessing includes denoising the CSI raw data by using the singular spectrum analysis SSA algorithm, and converting the denoised CSI raw data into a spectrogram through the Hilbert transform HHT, and extracting the fall or pseudo-spectrogram from the spectrogram. Fall CSI data, as the dataset to be identified;

Step 3: Divide the CSI data set to be identified into two parts, one part is used as a training set and the other part is used as a test set;

Step 4: Import the training set in Step 3 into the deep convolutional neural network CNN for feature extraction training, and use the test set to test the trained neural network for fall recognition;

The deep convolutional neural network includes eight deconvolution layers, a ResNet backbone network, a fully connected layer, a SoftMax activation function and a cross-entropy loss function, and its structure includes a generation module, a feature learning module, and an output module:

The generation module is used to convert the CSI data into a feature map with a spatial coding mode;

A feature learning module is used to map spatially encoded patterns to features that recognize falls;

The output module is used to output corresponding recognition results according to the extracted features;

The backbone network includes ResNet, Inception and VGG network, the activation function includes SoftMax activation function, and the loss function includes cross entropy function.

Further, the SoftMax activation function is:

Among them, i represents the ith sample in the data set, K represents the number of samples in the data set, and represents the probability value of the ith sample, and the sum of these probability values is 1.

Further, the cross entropy loss function is:

Among them, N represents the number of samples i, represents the label of sample i (positive class is 1, negative class is 0), represents the probability that sample i is predicted to be positive class;

Step 5: Collect the CSI raw data of the wireless signal in the detection area in real time, and perform the preprocessing of step 2, import the data set to be identified into the neural network in step 4 for processing, and determine whether the target falls in the detection area.

The above descriptions are only the embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied to other related technologies Fields are similarly included in the scope of patent protection of the present invention.

Claims (9)

1. a fall detection method based on wireless signal, is characterized in that, this method comprises the steps: Step 1: collect the channel state information in the wireless signal, and obtain the original CSI data; Step 2: preprocessing the CSI original data to obtain a data set to be identified; Step 3: Divide the CSI data set to be identified into two parts, one part is used as a training set and the other part is used as a test set; Step 4: Import the training set in Step 3 into the deep neural network for feature extraction training, and use the test set to perform a fall recognition test on the trained neural network; Step 5: Collect the CSI raw data of the wireless signal in the detection area in real time, and perform the preprocessing of step 2, import the data set to be identified into the neural network in step 4 for processing, and determine whether the target falls in the detection area. 2 . The wireless signal-based fall detection method according to claim 1 , wherein, in the step 1, the CSI original data includes CSI amplitude data. 3 . 3. The wireless signal-based fall detection method according to claim 1 or 2, wherein in the step 2, the CSI raw data is preprocessed to obtain a data set to be identified, including adopting singular spectrum analysis (SSA) The algorithm denoises the CSI raw data, and converts the denoised CSI raw data into a spectrogram through the Hilbert transform HHT, and extracts the fallen or pseudo-fallen CSI data from the spectrogram as the data to be identified. set. 4. The wireless signal-based fall detection method according to claim 1 or 2, wherein in the step 4, the deep neural network comprises a convolutional neural network CNN. 5. The wireless signal-based fall detection method according to claim 1 or 2, wherein the structure of the deep neural network comprises a generation module, a feature learning module and an output module: The generation module is used to convert the CSI data into a feature map with a spatial coding mode; A feature learning module is used to map spatially encoded patterns to features that recognize falls; The output module is used to output corresponding recognition results according to the extracted features. 6. The wireless signal-based fall detection method according to claim 4, wherein the convolutional neural network CNN comprises eight deconvolution layers, a backbone network, a fully connected layer, an activation function and a loss function . 7 . The wireless signal-based fall detection method according to claim 6 , wherein the backbone network comprises ResNet, Inception and VGG networks, the activation function comprises a SoftMax activation function, and the loss function comprises a cross-entropy function. 8 . 8. The fall detection method based on wireless signal according to claim 7, is characterized in that, described SoftMax activation function is:

Among them, i represents the ith sample in the data set, K represents the number of samples in the data set, σ _i represents the probability value of the ith sample, and the sum of these probability values is 1. 9. The wireless signal-based fall detection method according to claim 7, wherein the cross-entropy loss function is:

Among them, N represents the number of sample i, _yi represents the label of sample i, positive class is 1, negative class is 0, pi represents the probability that sample _i is predicted to be positive class.

Images