Empowering Consumer Electronics with Lightweight AI Application for Early Alzheimer’s Detection assisted by Acoustic Signals

Alzheimer’s disease (AD) has become a global challenge which requires early detection to alleviate suffering and to minimize the cost borne by the individuals and the public-private healthcare systems. In this context, the consumer electronics are being adapted and optimized for versatile healthcare systems fueled by the emergence of artificial intelligence (AI), edge computing and industry 4.0 standards. In this study, we explore machine learning models for AD detection, with client-server interaction for inference. We investigate the potential of acoustic features like frequency, intensity, spectral and temporal, as speech pattern recognition under cognitive decline. Our key areas include performance evaluation, effect of audio segmentation and data size, and analysis of variance (ANOVA) as feature selection technique. We present ANOVA-selected acoustic feature dataset to reduce computational complexity while maintaining high sensitivity and accuracy in a lightweight AI framework. The evaluation results demonstrate that the proposed model effectively detects an average of 17% decline of cognitive abilities among AD patients, achieving up to 95.3% correct classification rate. With a minimal complexity, the proposed framework offers an industry-ready AI solution for cognitive screening on both cloud and edge using Scikit-learn and Tensor frameworks for seamless integration into consumer electronics, wearables, and telemedicine platforms.