Vocal folds paralysis clasiffication using FLDA and PCA algorithms suported by an adapted block matching algorithm

Movement study in vocal folds recordings is basic to detect pathologies related to the movement, and specially the vocal folds paralysis. This approach involves four process steps: 1) a preprocessing stage 2) the analysis of the image textures applying Gabor filtering for the segmentation of the glottal area, 3) an adapted block matching algorithm using the Exhaustive Search method, and 4) classification using FLDA (Fisher's Linear Discriminant Analysis) and PCA (Principal Component Analysis) techniques. The adaptation of the block matching algorithm is made due to the heterogeneous nature of the ROI of each frame of the video sequence. The results show that our proposal works correctly to detect automatically vocal folds with paralysis and to distinguish them from healthy or pathological vocal folds with accuracy over the 95%. There is also shown the classification of the correct pathology over the 65% of the cases.