Jose Maria Perez-Macias

Snoring Detection using Emfit

snoring-project-image

Overview

This project focuses on developing algorithms and methods to detect snoring sounds during sleep. The project is based on the Emfit sensor, which is a contactless sensor that measures heart rate, breathing rate, and movement during sleep. The goal is to use the data collected by the sensor to detect snoring sounds and provide feedback to the user.

Background

Conventional snoring detection relies on microphones; a non-contact mattress sensor approach can improve comfort and reduce setup complexity.

Aim

To accurately identify snoring events from EMFiT signals, enhancing sleep analysis and user feedback.

Methods

Employing spectral analysis, machine learning (SVM, CNN), and source separation (NMF) to isolate and characterize snoring signals from raw sensor data.

Results

Experimental results show effective snoring detection, distinguishing snoring from normal breathing with high accuracy and consistency.

Resources

The project has produced the following articles

  1. Spectral analysis of snoring events from an Emfit mattress. JM Perez-Macias, J Viik, A Varri, SL Himanen, M TenhunenPhysiological measurement 37 (12), 2130 3 2016
  2. Snoring detection with emfit sleep mattress. JM Perez-Macias, SL Himanen, J Viik, M TenhunenJOURNAL OF SLEEP RESEARCH 25, 159-160 2016
  3. Assessment of support vector machines and convolutional neural networks to detect snoring using Emfit mattress
    Jose M. Perez-Macias, Sharath Adavanne, Jari Viik, Alpo Värri, Sari-Leena Himanen, and Mirja Tenhunen, The Engineering in Medicine and Biology Conference (EMBC 2017) (old draft here)
  4. Detection of snores using source separation on an Emfit signal . JM Perez-Macias, M Tenhunen, A Värri, SL Himanen, J Viik. IEEE Journal of Biomedical and Health Informatics