A novel recognition system for human activity based on wavelet packet and support vector machine optimized by improved adaptive genetic algorithm

A new human activities recognition system based on support vector machine (SVM) optimized by improved adaptive genetic algorithm (IAGA) and wavelet packet is proposed. Wavelet packet transform (WPT) is applied to extract the signatures from various actions. SVM is a powerful tool for solving the classification problem with small sampling, nonlinearity and high dimension. Genetic algorithm (GA) is employed to determine the two optimal parameters for SVM with highest predictive accuracy and generalization ability. Moreover, the IAGA adopts the dynamic cross rate and mutation rate according to the group fitness, thus effectively avoiding the disadvantages of the standard GA, such as premature convergence and low robustness. The average recognition accuracy rate goes up to 97.6%. In addition, the result of suggested method is also compared with other feature extraction methods which further demonstrate the superiority of WPT and generalization ability of IAGA. The aforementioned results clearly demonstrate that the proposed method is superior to the traditional method in activity recognition. (C) 2014 Elsevier B.V. All rights reserved.