Multivariate linear regression based activity recognition and classification

In this paper various activities such as sitting, standing and walking are recognized and classified using multivariate linear regression algorithm. Accuracy of recognition has been found to vary in a free living environment compared to the laboratory conditions. Hence the proposed work has considered free living environment. Zephyr BioHarness which has 3-axis accelerometer sensor is used for data acquisition. The proposed algorithm has been simulated in MATLAB. From the results obtained the accuracy of sitting activity is found to be 99.5 ± 0.3%, for standing activity 99.6 ± 0.4% and for walking activity 96.4 ± 0.2%. This shows more accuracy of 6%,7%,4%for sitting, standing and walking activities with the work taken for comparison.