Detection of Sleep Apnea and its Intensity in Adults

The following paper introduces a new method for identifying Obstructive Sleep Apnea (OSA), a widespread sleep disorder that impacts a large number of people globally. OSA is characterized by breathing pauses lasting from a few seconds to a minute or more. Our proposed approach utilizes audio signals for OSA detection. Existing studies require the use of ECG or EEG signals, which entail bulky equipment, electrodes, and instruments attached to the patient, resulting in a time-consuming and inconvenient signal extraction process. Conversely, our study uses audio signals due to their accessibility and convenience. To accurately detect OSA, we convert audio signals to time and frequency domains using FFT and DWT. Features are then extracted and used in the ANN model to obtain high accuracy and specificity in OSA detection. The proposed approach achieves high accuracy and specificity in detecting OSA. With the ANN model, we achieved an accuracy of 94.1%, sensitivity of 98.5%, and specificity of 88.7%. This indicates the potential of using audio signals for OSA detection, serving as a non-invasive and cost-effective method for OSA diagnosis.