Gait Pattern Recognition Based on Plantar Pressure Signals and Acceleration Signals
Meiyan Zhang,Dan Liu,Qisong Wang,Boqi Zhao,Ou Bai,Jinwei Sun
DOI: https://doi.org/10.1109/tim.2022.3204088
IF: 5.6
2022-01-01
IEEE Transactions on Instrumentation and Measurement
Abstract:Gait has been widely used in the fields of elderly care, posture correction, and identity recognition. By analyzing and processing the motion parameters and attitude data collected by the sensor, the current gait was determined. In order to pointedly prevent falling in daily life, we proposed a gait pattern classification method based on multisensor. Falling and the gait patterns (standing, sitting/rising, squatting/rising, walking, and running) that falling is likely to happen in daily life were distinguished. Besides, we integrated pressure signals with acceleration signals to compensate for the insufficient data provided by single sensor that cannot fully reflect the complex human motion and solved the problem that falling detection based on acceleration signals is prone to misclassify because certain postures are of similar acceleration changes with falling. For further data analysis, the collected gait data were then transmitted to upper machine for signal processing through the designed wireless network. Combined with the characteristics of gait patterns, we analyzed the corresponding pressure signals, acceleration signals, and resultant acceleration signals. Wavelet energy entropy features and wavelet packet energy features were subsequently extracted from the collected gait data. Finally, we input the randomly selected test data into the established extreme learning machine (ELM) and $K$ -nearest neighbor (KNN) model to test gait pattern recognition effects. The performance of ELM algorithm was better in terms of processing time and classification results, with the highest average identification accuracy, precision, and recall rate of 0.974, 0.937, and 0.936, respectively. Besides, precision–recall (PR) curve was optimal, with the largest area of 0.973. Our presented algorithm responded rapidly and prevented falling in daily lifetime, which can be applied to health monitoring systems to detect daily activities of the elderly promptly.