Mattela, Govardhan and Tripathi, Manmohan and Pal, Chandrajit
(2022)
A Novel Approach in WiFi CSI-Based Fall Detection.
SN Computer Science, 3 (3).
ISSN 2662-995X
Full text not available from this repository.
(
Request a copy)
Abstract
Falls are unforeseen events, that can potentially be injurious or even fatal, happen either because of health-related issues like weakness, faint ing, abnormal blood pressure, or external factors like slipping and tripping. According to the National Safety Council, falls are one of the leading causes of unintentional injuries accounting for 0.6 million fatal falls every year globally. This necessitates an efficient fall detection mechanism that can automatically detect falls and raise the alarm. Existing methodologies including wireless body-worn sensors do exist for the fall detection task, however at the cost of discomfort arising out of the multiple sensors for the patient wearing it. In this paper, we have explored the usage of the pattern of the data stream from channel state information (CSI) extracted from the Wi-Fi signals received from a single antenna domestic router, which can recognize a human fall activity in real-time and can raise an alarm, without the need of any body-worn sensor. This is done by employing Artificial Intelligence to build deep learning models to classify the extracted features from the CSI data stream. We have designed and implemented predictive deep neural network sequential models like Long Short-Term Memory (LSTM) and lightweight autoencoders for fall detection with extremely promising performance and accuracy ranging between 97 to 99% approximation in three different indoor scenarios with varied layout topology of transmitter–receiver links. This technology can prove to be a lifesaver in situations when a person is unable to raise any medical alert after being fallen. © 2022, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.
Actions (login required)
 |
View Item |
Altmetric
Altmetric