1. according to the invention, by collecting the facial feature information of the detected person, the generated signal is decomposed into a plurality of sub-signals by adopting a variational self-coding independent component decomposition principle, the heart rate signals which do not meet the requirements in the sub-signals are eliminated, the heart rate signals which meet the requirements are integrated, a heart rate curve with more accurate representation can be obtained, detailed and accurate heart rate indexes can be obtained from the heart rate curve, and the detection accuracy is higher.

2. The invention provides a non-contact heart rate detection method, which reduces the complicated wearing process and excessive exposure contact, enables the detection process of a patient to be more efficient, and reduces the risk of infection of medical staff.

Drawings

FIG. 1 is a schematic diagram of signal decomposition according to an embodiment of the present invention;

fig. 2 is a schematic diagram of a variational self-encoder according to an embodiment of the present invention.

Detailed Description

The embodiments of the present invention will be further described with reference to the accompanying drawings.

Example (b):

s1, inputting facial feature data of the detected person;

s2, extracting the facial feature information of the detected person;

s6, integrating the heart rate signals and establishing a heart rate curve;

In another embodiment, as shown in fig. 1, in step S4, the method for decomposing the signal information is as follows:

s41, inputting a time domain signal F (0) (F (0) is a time domain signal, mixing a plurality of signals together, wherein the signals comprise useful signals and useless signals, performing Fast Fourier Transform (FFT) on the signals, and converting the signals into frequency domain signals;

s44, outputting the difference between the time-domain signal F (0) and the sub-signal F (1), and obtaining a new signal F (1), i.e. F (1) ═ F (0) -F (1);

s45, inputting the signal F (1), repeating steps S41-S43 to obtain a sub-signal F (2), and outputting the difference between the time-domain signal F (1) and the sub-signal F (2) to obtain a new signal F (2), i.e. F (2) ═ F (1) -F (2);

s46, inputting a new signal F (2), and repeating the steps S41-S45 until n sub-signals are output.

wherein, F (0) is the original signal, and F (i) is the ith sub-signal.

As shown in the figure2, the masking neural network adopted in this embodiment combines the neural network with the independent component analysis, so that an effective mask can be directly output from the input signal, wherein the gradient equation is:

the above-mentioned embodiments only express the specific embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.

Claims

1. A heart rate detection method based on a mask neural network independent component decomposition principle is characterized by comprising the following steps:

s1, inputting facial feature data of the detected person;

s2, extracting the facial feature information of the detected person;

s3, analyzing the facial feature information of the detected person, screening out key facial feature information, and outputting a signal of the key facial feature;

s4, decomposing the signal and outputting a plurality of sub-signals;

s6, integrating the heart rate signals and establishing a heart rate curve;

2. The method for detecting heart rate based on the mask neural network independent component decomposition principle as claimed in claim 1, wherein in step S4, the method for decomposing the signal information is as follows:

s41, inputting a time domain signal, performing fast Fourier transform, and converting the time domain signal into a frequency domain signal;

s42, inputting the frequency domain signal into a mask neural network, acquiring a frequency mask, multiplying the frequency mask and the frequency domain signal, and outputting a filtered frequency domain signal;

s43, inputting the filtered frequency domain signal, performing inverse fast Fourier transform, and reducing the frequency domain signal into a time domain signal to obtain a sub-signal;

s44, carrying out difference on the original time domain signal and the sub-signal obtained in the S43 to obtain a new signal;

s45, inputting a new signal, repeating the steps S41-S44 and outputting a plurality of sub-signals.

3. The method for detecting heart rate based on the mask neural network independent component decomposition principle as claimed in claim 2, wherein in step S44, when the difference between the time domain signal and the sub-signal is calculated, the signal loss is substituted, and the calculation formula of the signal loss is as follows:

wherein, F (0) is the original signal, and F (i) is the ith sub-signal.