Multi-sensor signal fusion method based on deep learning for gait classification

Abstract

The invention belongs to the technical field of biological feature recognition, and particularly relates to a multi-sensor signal fusion method based on deep learning for gait classification. Abnormalgaits are classified by constructing a deep neural network, and multi-source heterogeneous information source data from an IMU inertial sensing unit and SEMG surface electromyography are fused by utilizing a convolutional neural network; The fusion content comprises a data layer (CNN input layer), a feature layer (CNN pooling layer 1 to convolutional layer 2) and a decision layer (CNN output layer) fusion, so that multi-source heterogeneous sensor information is completely extracted, the classification precision of the classifier is improved, the data preprocessing workload is reduced, and the classification accuracy and the judgment efficiency are improved. It is verified that the classification effect of the method in multiple abnormal gait classification tasks is remarkably improved compared with that of a single-mode sensor, and in an abnormal gait six-classification task mentioned in the embodiment, the classification accuracy reaches 99.15%, and is improved by about three percentage points compared with that of a single IMU information source CNN network.

Description

technical field [0001] The invention belongs to the technical field of biological signal sensing, and in particular relates to a multi-sensor signal fusion method for gait classification. Background technique [0002] Gait refers to the posture shown when people walk, and is one of the important biological characteristics of the human body. Abnormal gait is mostly related to the lesion. As an important feature reflecting the health status and behavior ability of the human body, accurate and credible gait information can be obtained in time, and the abnormal gait classifier can be trained to give timely early warning of abnormal gait, and it can be monitored for a long time. Monitoring and evaluation have important guiding significance in medical diagnosis and disease prevention. [0003] Multi-sensor data fusion MSDF (Multi-sensor Data Fusion) technology was first widely used in the military. Recently, with the increasing maturity of biomedical information collection techno...

AI Technical Summary

Problems solved by technology

Based on sensors, there are many single-sensor solutions based on lower limb movement information, based on lower limb surface electromyography, and based on plantar pressure distribution. However, compared with multiple sensors, the data of a single sensor is thin, and it cannot obtain complete and comprehensive information about the object.

Images