Enhanced Auditory Experience for Hard-of-Hearing People in Multi-Speaker Environments through Selective Sound Amplification

Abstract

Abstract Hearing loss remains a significant global health concern, affecting communication, quality of life, and cognitive function of individuals. Traditional hearing aids, though clinically effective, often remain underutilized due to high cost, limited accessibility, social stigma, and their inadequacy in handling real-world listening environments. These devices typically offer general sound amplification without the ability to distinguish between relevant and irrelevant audio sources. As a result, users frequently experience difficulty engaging in conversations within multi-speaker or noisy settings, leading to frustration and abandonment of assistive technology. Existing mobile applications, although more affordable and accessible, largely replicate this indiscriminate amplification approach and provide only basic noise suppression. This research addresses these shortcomings by proposing a smartphone-based hearing aid application that integrates selective sound amplification with advanced noise reduction. The system utilizes speaker diarization and identification technologies to isolate and enhance the voice of a desired speaker while suppressing background noise and competing voices. Deep learning-based models for voice activity detection and real-time noise suppression are incorporated to optimize intelligibility and listening comfort. In parallel, a user-centric interface was developed to enable intuitive configuration of amplification preferences, facilitating ease of use among individuals with limited technical experience. The study employed a Systems Development Research approach, combining empirical investigation with iterative system design and evaluation. The research context focused primarily on university students with mild to moderate hearing impairment, a population chosen for their frequent exposure to challenging auditory environments such as classrooms and lecture halls. Participants were recruited from the Centre for Disability Research, Education and Practice, University of Colombo, and their experiences, challenges, and feedback directly informed system development. The application was evaluated through acoustic measurements and structured user studies, incorporating quantitative ratings and qualitative interviews to gather comprehensive insights. Results demonstrate substantial improvements in speech