Automatic Speech Analysis of Conversations for Dementia Detection Using LSTM and GRU

Abstract

Neurodegenerative diseases, such as dementia, can impact speech, language, and the capability of communication. A recent study to improve the dementia detection accuracy studied the usage of CA (Conversation Analysis) of interviews among patients and neurologists to distinguish among progressive Neurodegenerative Memory Disorders patients & those with (non-progressive) Functional Memory Disorders (FMD). However, manual CA is costly for routine clinical use and difficult to scale. In this work, we present an early dementia detection system using speech recognition and analysis based on NLP technique and acoustic feature processing technique apply on multiple feature extraction and learning using a LSTM (Long Short-Term Memory) and GRU which remarkably captures the temporal features and long-term dependencies from historical data to prove the capabilities of sequence models over a feed-forward neural network in forecasting speech analysis related problems.