Health status monitoring system based on speech analysis

Abstract

Abstract In recent years, the health of the elderly has become the focus of medical treatment. In daily life, the physical condition of the elderly cannot be fed back to their families in time, and the health level of the elderly at home cannot be accurately judged. In order to solve the above problems, we put forward a method to monitor the health of the elderly with speech recognition system. This method can not only monitor the health of the elderly in real time, but also inform the family members in time when there is something wrong with their health. The specific steps are as follows: firstly, sound waves are converted into electrical signals by sound sensors. Because the frequency band distribution of the input sound wave signals is complex, we choose band-pass filters to shield other noises, and choose the sound frequency band belonging to specific elderly people as input signals. Because of the particularity of sound wave signals, they are time-varying signals, which need to be processed. We extract characteristic parameters and frame-by-frame stable digital signals to obtain stable signals for subsequent processing. The filter and signal processing chip are designed by FPGA, and the finished product is prepared after layout drawing. In the process of protecting the health of the elderly, the invention is equipped with image recognition, which is convenient for timely monitoring the living state of the elderly and whether there is sudden disease. Comprehensive speech recognition chip is used to protect the healthy life of the elderly. template matching Speech endpoint detection Speech database , Feature parameter extraction - - Template matching Judging whether the elderly are healthy or not Computational recognition Figure -Speech in training set (verification) Test set - : la oc Outside the training set Speech with noisy Voice (test) background Unrecognized speech Figure2