Implementation of a Computational Intelligence System to Identify Respiratory Distress from Digital Signal Processing of Voice

This paper presents a computational intelligence system that allows us to identify people with respiratory distress in an automatic and non-invasive way. This is accomplished through the digital signal processing techniques of voice, integrating the calculation of acoustic, spectral, temporal, and sta-tistical parameters and neural networks. The computational method is implemented in a client-server architecture that allows recording two sustained vowel sounds (vowel “a” and “o”) of a spanish-speaking population group aged between 18 and 49 years old. Results show that the vowel “a” is more efficient in women with a hit rate of 97.62% for detecting respiratory distress. In contrast, the vowel “o” is more efficient in men, reaching a hit rate of 96.77%. This system is expected to contribute to the development of novel support tools that could be used in the medical field to facilitate the diagnosis of respiratory diseases and prevent their spread.