| CPC A61B 5/02416 (2013.01) [A61B 5/721 (2013.01); A61B 5/7257 (2013.01); A61B 5/7264 (2013.01); A61B 5/742 (2013.01); A61B 2562/0219 (2013.01)] | 20 Claims |

|
1. A vital sign detection method, comprising:
obtaining an acceleration signal and photoplethysmographic (PPG) signals of a plurality of channels within a preset time period;
calculating a correlation coefficient matrix among the PPG signals of the plurality of channels, wherein at least one correlation coefficient in the correlation coefficient matrix is less than or equal to a first threshold;
calculating a variance of the PPG signals of each channel in the plurality of channels;
determining that the variances of the PPG signals of all the channels are greater than or equal to a second threshold;
performing frequency domain fusion after performing a fast Fourier transformation on the PPG signals of the plurality of channels, to obtain a frequency domain signal of a fused PPG signal;
performing the fast Fourier transformation on the acceleration signal to obtain a frequency domain signal of the acceleration signal;
inputting the frequency domain signal of the fused PPG signal and the frequency domain signal of the acceleration signal into a deep sequence neural network to obtain a value of a vital sign, wherein the vital sign comprises a heart rate, wherein inputting the frequency domain signal of the fused PPG signal and the frequency domain signal of the acceleration signal into the deep sequence neural network to obtain the value of the vital sign comprises:
splicing the frequency domain signal of the fused PPG signal and the frequency domain signal of the acceleration signal into an eigenvector, and inputting the eigenvector to a first fully connected layer of the deep sequence neural network;
outputting, by the first fully connected layer to a normalization layer, a non-linearized eigenvector to a normalization layer of the deep sequence neural network, wherein neurons in the first fully connected layer are fully connected to neurons in the normalization layer;
normalizing, by the normalization layer, the non-linearized eigenvector, and outputting the normalized eigenvector to a recurrent network layer of the deep sequence neural network;
accumulating, by the recurrent network layer, time sequences of normalized eigenvectors, and outputting the time sequences to a second fully connected layer of the deep sequence neural network, wherein neurons in the second fully connected layer are fully connected to neurons in the recurrent network layer; and
outputting, by the second fully connected layer, the calculation result; and
displaying the value of the vital sign.
|