Body Falling Gesture Recognition Based on SOM and Triaxial Acceleration Information

Abstract

Abstract-In order to improve the performance of fall detection system for the elderly based on triaxial acceleration sensor, and accurately to judge the fall direction of human body, a method was put forward based on self-organizing map neural network (SOM) and the information of triaxial acceleration sensor to cluster and analyze the human motion. To verify the recognition results of the SOM method, 130 samples of 13 common action including fall were participated in the SOM network testing. The results show that the sensitivity, specificity and accuracy of the new system were 90%, 96.7%, 94.6%, respectively. These results were better than those of the method of threshold value.

Landmark map: An extension of the self-organizing map for a user-intended nonlinear projection

“…For instance, they have been used to classify electroencephalography (EEG) signals in applications of human-computer interaction (HCI) such a brain-computer interface (BCI) [12]. Furthermore, SOMs have also been utilized for facial recognition [13], hand and body tracking [14], gesture recognition [15], fault detection [16], mental task classification of EEG for BCI [17], EEG-based emotion recognition [18], emotion recognition using geometric facial features [19], and hazard detection of motorcycles [20]. Although SOM is useful for mining unlabeled data because of its unsupervised learning, they do not take advantage of labeled or partly labeled data in supervised or semi-supervised learning.…”