On the possibility of using a hybrid approach in the recognition of potentially dangerous physiological conditions

The paper is devoted to the recognition of potentially dangerous physiological conditions. It is proposed to convert one-dimensional signals used in medical diagnostics into a video sequence. To recognize a video sequence, it is proposed to combine a convolutional and recurrent neural network. The effectiveness of multiclass classification of various physiological conditions is compared using a method combining recurrent and convolutional neural networks (accuracy metric is 0.98) with recurrent (0.53) and convolutional neural networks (0.41). The high efficiency of the proposed approach is shown.