Walking Gait Phase Detection Based on Acceleration Signals Using Voting-Weighted Integrated Neural Network

Abstract

Human gait phase recognition is a significant technology for rehabilitation training robot, human disease diagnosis, artificial prosthesis, and so on. The efficient design of the recognition method for gait information is the key issue in the current gait phase division and eigenvalues extraction research. In this paper, a novel voting-weighted integrated neural network (VWI-DNN) is proposed to detect different gait phases from multidimensional acceleration signals. More specifically, it first employs a gait information acquisition system to collect different IMU sensors data fixed on the human lower limb. Then, with dimensionality reduction and four-phase division preprocessing, key features are selected and merged as unified vectors to learn common and domain knowledge in time domain. Next, multiple refined DNNs are transferred to design a multistream integrated neural network, which utilizes the mixture-granularity information to exploit high-dimensional feature representative. Finally, a voting-weighted function is developed to fuse different submodels as a unified representation for distinguishing small discrepancy among different gait phases. The end-to-end implementation of the VWI-DNN model is fine-tuned by the loss optimization of gradient back-propagation. Experimental results demonstrate the outperforming performance of the proposed method with higher classification accuracy compared with the other methods, of which classification accuracy and macro-F1 is up to 99.5%. More discussions are provided to indicate the potential applications in combination with other works.

A Review of Gait Phase Detection Algorithms for Lower Limb Prostheses

“…The models are adopted depending on the type of sensors that are installed for recording the signals of the gait. Nowadays, wearable sensors are widely used for gait phase recognition systems: Wearable force-based measurements [ 9 , 21 , 26 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 ], Electromyographic (EMG) sensors [ 55 , 56 ], Inertial Measurement Units (IMUs) [ 9 , 19 , 29 , 41 , 57 , 58 ], and joint angular sensors [ 24 , 59 , 60 , 61 , 62 ] are used specifically for the detection of the gait. The studies showed that the methods that used force-based measurements such as Force Sensing Resistors (FSRs), footswitches, and foot pressure insoles yield the highest precision for detection [ 7 ].…”

An Acceleration Based Fusion of Multiple Spatiotemporal Networks for Gait Phase Detection

“…Studies have shown that the swing phase segment accounts for about 40% of the entire gait cycle and the standing phase accounts for about 60% of the entire gait cycle. According to the previous analysis [1], the schematic diagram of gait cycle division is shown in Figure 5. 60% of the entire gait cycle.…”

“…Among them, lower extremity exoskeleton has important research value in the medical field, its main potential is to enhance the patient’s ability to move in rehabilitation therapy, and enhance physical function after receiving treatment, and hope to improve their quality of life as much as possible. Among them, gait recognition technology is an important technical guarantee for the robot to process a large amount of instantaneous time series data, which is one of the most important features to display the posture and phase of each specific patient [ 1 ]. Therefore, there is an urgent need to accurately judge the gait phase of the human lower extremity state change in order to enhance the consistency and coordination of human-computer interaction [ 2 ].…”

“…According to previous studies, gait-phase-recognition methods can generally be divided into two categories. The first type is the threshold method, which determines the corresponding phase information by setting the corresponding threshold [ 1 ]. However, this type of algorithm is too rough and difficult to deal with complicated situations.…”

“…Many researchers place the IMU on the instep, calf and thigh. For example, Yan et al [ 1 ] designed a voting-weighted integrated deep learning algorithm, and by inputting acceleration signals on the instep, calf and thigh into the model, successfully detected four subphases of pedestrians, and achieved very good results. Identify the effect.…”