Fall detection approach based on combined two-channel body activity classification for innovative indoor environment.

Autor: De, Anurag, Saha, Ashim, Kumar, Praveen, Pal, Gautam
Zdroj: Journal of Ambient Intelligence & Humanized Computing; Sep2023, Vol. 14 Issue 9, p11407-11418, 12p
Abstrakt: Human fall detection plays a vital role in monitoring senior citizens safely while being alone. In recent works, vision-based techniques provide favorable and effective results. In this paper, a combined two-channel fall detection approach is proposed using Support Vector Machine (SVM) and K-Nearest Neighbors (K-NN) classification models based on the displacement of significant spatial features of the foreground image. Initially, training of both fall and daily activity scenarios is done using a standard fall detection dataset. Keyframes consisting of significant body shape features are then obtained from the surveillance video subjected to the two-channel classification model. We consider the classification results if both the channels generate similar outputs, failing which, additional intelligence is used to classify the fall and daily activity event. The keyframe selection is based on the displacement in height-to-width ratio and displacement in horizontal and vertical centroid movement of the object having a threshold higher than a preset value. The proposed fall detection system achieves a peak accuracy of 98.6% and sensitivity of 100% in detecting falls. The proposed model achieves satisfactory performance in comparison to existing state-of-the-art techniques. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: