Surface roughness discrimination using unsupervised machine learning algorithms

In this paper, the ability of unsupervised surface roughness discrimination is explored based on the developed bioinspired artificial fingertip. At first, the original signals are analyzed and discriminated with the most widely used unsupervised algorithm, Kmeans clustering, applied. Then the technique of discrete wavelet transform and algorithm of sequential forward selection are utilized successively to select the most discriminative feature combination. The unsupervised discrimination results are presented and compared by using Kmeans based on different distances. The highest test accuracy reaches 72.93%+/- 12.55% when the algorithm of Kmeans-SEuclidean is adopted and six discriminative features are selected, which showed that the developed tactile fingertip is effective in discriminating surface roughness based on unsupervised learning.