AI-Driven Audio Signal Analysis For Healthcare Monitoring

The proposed cough detection system presents an effective approach for automated respiratory sound analysis using machine learning and deep learning techniques. The system is designed to classify audio inputs into cough and non-cough categories by leveraging multiple classification models, including Logistic Regression, Random Forest, and Convolutional Neural Networks (CNN). The workflow includes essential stages such as audio preprocessing, noise reduction, normalization, and feature extraction using Mel-Frequency Cepstral Coefficients (MFCC) and spectrogram representations. The comparative analysis of models demonstrates that traditional machine learning algorithms such as Logistic Regression and Random Forest perform reasonably well on engineered features; however, they are limited in capturing complex temporal and spatial patterns present in audio data. In contrast, the CNN model achieves superior performance by automatically learning hierarchical feature representations from spectrogram images, making it more suitable for audio-based classification tasks. Furthermore, the system is capable of handling both offline dataset evaluation and real-time audio input, which enhances its practical usability in real-world health monitoring scenarios. The results indicate that AI-based audio classification systems can significantly contribute to early detection and monitoring of respiratory conditions, thereby supporting healthcare applications. Despite the promising results, the system still has limitations such as dependency on dataset quality, sensitivity to background noise, and limited generalization across diverse real-world conditions. These challenges highlight the need for more robust training datasets and improved model architectures. Overall, this work establishes a strong foundation for intelligent cough detection systems and demonstrates the potential of deep learning in healthcare audio analytics. Future enhancements such as large-scale dataset integration, lightweight model deployment, and mobile-based real-time monitoring systems can further improve the applicability and efficiency of the proposed approach. Overall, the proposed system demonstrates that machine learning and deep learning techniques can effectively be applied to audio-based healthcare applications with high reliability. This work can serve as a foundation for developing more advanced and real-time intelligent health monitoring systems in the future.

1. T. Giannakopoulos and A. Pikrakis, Introduction to Audio Analysis: A MATLAB Approach, Academic Press, 2014.   2. B. McFee et al., “librosa: Audio and Music Signal Analysis in Python,” Proceedings of the 14th Python in Science Conference, 2015.   3. I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning, MIT Press, 2016. 4. S. B. Davis and P. Mermelstein, “Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences,” IEEE Transactions on Acoustics, Speech, and Signal Processing, 1980.   5. L. Breiman, “Random Forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001.   6. K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” ICLR, 2015.   7. T. Heittola et al., “Audio Set and Audio Event Detection,” IEEE ICASSP, 2018.   8.Scikit-learn Developers, “Scikit-learn: Machine Learning in Python,” https://scikit-learn.org