Alzheimer’s Detection with Speech Using Singular Spectrum Analysis Trend Extraction Method

Abstract

With the aging of the population in various countries, the impact of Alzheimer’s disease on humans is becoming more and more obvious. It is very necessary to propose an Alzheimer’s detection system. This paper attempts to use the singular spectrum analysis trend extraction method to complete the task of Alzheimer’s detection with speech. First, the singular spectrum analysis is performed on the speech signal, and the components obtained by the singular spectrum analysis are divided into a trend part and a detrend part according to the energy ratio. Second, feature extraction is performed on the trend part and the detrend part of the speech signal respectively. These features include multidimensional voice program, the gammatone frequency cepstral coefficient, and the Power-normalized cepstral coefficients. Then, use random forest to calculate the importance of feature vectors, and select the top 30 features that random forest considers the most important as the features used in this article. Finally, random forest is used for classification.