It was found that all 8 input features produced a maximum LOO accuracy of 94.2% using a SVM with Gaussian kernel, s=0.5 and C=10. Due to space constraints, we report only the best accuracies over the range of SVM kernels (Table 2) where on average, the Gaussian kernel was found to provide better classification compared to the polynomial and linear kernels. The SVM was better overall at detecting the healthy subjects compared to the OA gait as depicted by the higher specificities.