A Non-Contact Paraparesis Detection Technique Based on 1D-CNN

In clinical practice, doctors are using bedside tests to assist in the diagnosis of paraparesis. The disadvantage is that it depends on the doctor’s clinical experience and the supervisor’s judgment. Therefore, there is an urgent need for an objective and efficient diagnostic equipment. With the rapid development of wireless technology, ubiquitous RF signals become a promising sensing technology. In this study, we propose a non-contact wireless sensing method based on RF signals to detect paraparesis. Our system can reduce the burden on doctors and improve work efficiency. Outlier filters and wavelet hard threshold decomposition are used to filter the wireless signal. A 1D-CNN model is designed to automatically extract valid features and classifications. The results analyze in two bedside tests, our system perform efficiently and accurately patient screening with suspected paraparesis. This provide more effective guidance and assistance for further treatment. The proposed method has an average accuracy of 99.4% and 98.5% in the Barre test and Mingazzini test respectively.