Machine-learning-based children’s pathological gait classification with low-cost gait-recognition system

Table 2 Experimental results data of dynamic section

Algorithm	Recall	Average precision (%)	Time cost (ms)	Accuracy/time cost P-value
LDA + SVMrbf	0.96/1.00/0.96/0.93	96.24	6.10 4.40 75.20 27.40 5.10	0.991 / < < 0.01
LDA + SVMlin	1.00/1.00/1.00/1.00	97.79		–
PCA + SVMlin	0.68/1.00/0.99/0.81	87.24		0.007 / < < 0.01
PCA + SVMrbf	0.90/1.00/1.00/0.86	94.13		0.107 / < < 0.01
LDA + NN	0.96/1.00/0.97/0.92	97.38		0.702 / 0.064

“–” means the P-value of this column is computed based on this algorithm. In the Recall column, the value separated by “/” means recall value of normal, toe-in, toe-out, and flat, respectively, from left to right. In the Accuracy/time cost P-value column, the value separated by “/” means P-value of prediction accuracy and P--value of time cost P--value between algorithm on this row and PCA + SVMlin, respectively, from left to right

ISSN: 1475-925X