Fall Detection for the Elderly using a Support Vector Machine
Patimakorn Jantaraprim1, Pornchai Phukpattaranont2, Chusak Limsakul3, Booncharoen Wongkittisuksa4
1Patimakorn Jantaraprim, Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
2Pornchai Phukpattaranont, Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
3Chusak Limsakul, Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
4Booncharoen Wongkittisuksa, Department of Electrical Engineering, Faculty of Engineering, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
Manuscript received on February 15, 2012. | Revised Manuscript received on February 20, 2012. | Manuscript published on March 05, 2012. | PP: 484-490 | Volume-2 Issue-1, March 2012. | Retrieval Number: F0359121611/2012©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: We propose a short time min-max feature for improving fall detection performance based on the specific signatures of critical phase fall signal, acquired using a tri-axial accelerometer on a torso. Our proposed feature has been validated by a Support Vector Machine with two-fold cross-validation. Fall and scripted activities were tested in the experiment. Performance was evaluated by comparing the short time min-max with a maximum peak feature. The results obtained from 420 sequences show that the performances of short time min-max feature can approach 98.2% sensitivity and 100% specificity for a radial basis function kernel, which are better than those from the maximum peak feature for all testing kernels. The short time min-max feature also uses one sensor for the body’s position without a fixed threshold for 100% sensitivity or specificity, and without additional processing of a posture after a fall. The simplicity and high performance of our proposed feature makes it suitable for implementation on a microcontroller for use in practical situations.
Keywords: Fall detection, Critical phase, Short time min-max feature, Support Vector Machine.