基于PSO模式搜索的跌倒检测算法研究

针对粒子群算法存在后期趋同性严重、收敛速度缓慢以及易陷入局部极小点等缺点，将模式搜索算法引入粒子群算法，对支持向量机参数进行优化，应用于跌倒检测中。首先，使用穿戴式设备收集跌倒检测数据集，将初始数据进行均值滤波以消除噪声的影响；然后，滤波后的数据特征提取，将提取的多维数据使用奇异值分解算法进行降维；最后，降维后的数据将用来检验粒子群模式搜索算法的优劣。通过与支持向量机算法和支持向量机算法加粒子群算法进行对比，粒子群模式搜索算法在跌倒检测中特异性和灵敏度都得到了提高。

Research on fall detection algorithm based on PSO pattern search

Aiming at the shortcomings of particle swarm optimization(PSO) such as serious convergence, slow convergence and easy to fall into local minimum, it introduced the pattern search algorithm into PSO to optimize the parameters of support vector machine(SVM) and applied to fall detection. Firstly, it used a wearable device to collect the fall to the detection data set, and filtered the initial data to eliminate the influence of noise. Secondly, it extracted the filtered data feature and reduced the dimensionality of the extracted multi-dimensional data by singular value decomposition algorithm(SVD). Finally, it used the dimensionality-reduced data to test the PSO pattern search algorithm. Compared with support vector machine(SVM) and support vector machine(SVM) plus particle swarm optimization(PSO), it improved the specificity and sensitivity of PSO in fall detection.