Embedded System for Early Sleep Apnea Detection Based on Mel-Frequency Cepstral Coefficient (MFCC) and Deep Learning Technique

Early Sleep Apnea Detection

My contributions began at the system concept design stage, specifically regarding the integration of machine learning and embedded systems for the early detection of sleep apnea. I was involved in searching for and evaluating snoring datasets from Kaggle, which were used as the basis for training a snoring sound classification model. Additionally, I handled the data preprocessing and feature extraction using MFCC as a representation of sound characteristics before inputting them into the machine learning model.

On the embedded system side, I contributed to hardware architecture design, sensor selection, device assembly, and firmware development on the ESP32 for data acquisition and collection from the health sensors used in the system. I also assisted with data communication integration between the embedded device and the main processing system.

In the machine learning section, I was involved in the experimentation and evaluation of several classification models, namely CNN, Random Forest, and SVM. Specifically, I performed training and hyperparameter tuning on the SVM model to find the configuration that provided the best performance. Additionally, I participated in analyzing the experimental results and comparing the performance of each model.