Acoustic analysis of voice for detection of speech disorder for amyotrophic lateral sclerosis

A method of acoustic signal analysis with sustain vowel phonation for detection of amyotrophic lateral sclerosis (ALS) is considered. A method for segmentation of the voice signal into periods of the fundamental tone, which is used for evaluation of the jitter and shimmer parameters, is proposed. A comparison of two ALS detectors was performed. The first detector was trained using voice features extracted by the proposed method, while the second detector was trained using features obtained with PRAAT toolkit. The result showed a significant improvement (by 20 %) in the accuracy of detecting ALS disease using the proposed method.