AN INTELLIGENT FRAMEWORK FOR MULTISOUND DETECTION AND ENHANCEMENT IN DIGITAL STETHOSCOPES

Author: Shreyaa V, Zameer Ali S, Pragadeesh R, Dr.Deepa D

Published On: 2025-03-25

DOI: https://doi.org/10.70127/irjedt.vol.7.issue03.1944

Technological advancements in digital signal processing and AI have enhanced auscultation-based diagnostics. Conventional stethoscopes function but suffer from ambient noise and expert understanding. The desire for smart and automated diagnostic systems resulted in digital stethoscopes that provide better analysis of heart and lung sounds. This research seeks to overcome the deficiencies of traditional auscultation by combining machine learning with digital stethoscopes. The system applies Gaussian noise filtering and spectral subtraction for the elimination of noise and NMF for separation of sound. Subsequently, heart and lung disease classification is conducted based on derived features using a CNN-LSTM. An LLM based on GPT-2 offers interactive explanations of findings for enhanced understanding. Experimental outcomes indicate 91.46% accuracy which is significantly higher than with traditional auscultation techniques. The system is able to distinguish between heart and lung sounds and can accurately diagnose. The AI driven explanations further enhance the systems application in clinical practice. These outcomes demonstrate the capability of AI driven digital stethoscopes in non-invasive diagnosis and decision making for medical professionals.

Key Words: Digital stethoscope, CNN-LSTM, sound separation, signal processing, disease classification, AI.