Doppler radar-based human breathing patterns classification using Support Vector Machine

Monitoring and recognizing human breathing patterns is of great importance in preliminary disease diagnosis. This paper presents a noncontact human breathing patterns classification method based on the Doppler radar. The proposed classification method can be suitable for discriminating the breathing patterns automatically. The Support Vector Machine (SVM) classifier, which solves the nonlinear problem using kernel function, is widely used in pattern recognition. It is selected to classify four typical breathing patterns. Three features from the time-domain and short-term energy-domain are extracted for the classification. In the experiment, the SVM classifiers with six different kernel functions have been tested on a dataset of 60 samples from five healthy subjects. Through the 10-fold cross-validation, experimental results show that the cubic SVM classifier has the best classification accuracy rate of 93.3%.