Novel Approaches to Activity Recognition based on Vector Autoregression and Wavelet Transforms

The recognition of daily activities has been a long-running research domain, which has received increasing attention over the past few years. This is due to the proliferation of personal devices which are capable of reporting the physical signals generated during these activities. Being a classification problem, the primary focus is on suitable modalities for feature extraction and proper choice of classifiers. In this work we investigate the performance of two novel approaches to feature extraction based on Vector Autoregression and Wavelet Transforms together with four different classifiers. The results indicate that the two proposed feature extraction methods are suitable for this domain. In addition, the Canonical Correlation Forests classifier has been found to be a promising candidate for inference in the domain of Activity Recognition.