Wireless Sensor and Smartphone-based System for Detection of Sleep Apnea

This study describes a real-time system to identify apnea or hypopnea events that occur during sleeping of the sleep apnea patient. The suggested solution uses a smart phone’s Bluetooth communication interface to transfer respiration data from a oronasal temperature sensor. A SVM-based binary classifier was built in the smartphone to categories a typical eight-second buffered/windowed respiratory signal into the normal (N) and disordered (D) breathing states. A high priority interrupt loop that interrupts every two seconds updates the buffer while also calling the feature calculation and classification subroutines during each interrupt cycle. In this way, the relevant period is recognised as an apnea/hypopnea event if the classifier’s five consecutive decisions are judged to be disordered (D). With the aid of the SVM-based Recursive Feature Elimination (RFE) technique, a set of seven features were screened from the initially selected set of 20 features. During the testing procedure, it was discovered that the suggested system’s event detection accuracy was 90.1%. The suggested smart phone-based system can also initiate an alarm when an event lasts longer than 10 seconds and can interact in real time with a remote computer using the global GPRS/UMTS wireless network.