Smartphone Based Human Activity Recognition Using RNN Variants
Sonia Abraham, Rekha K. James · 2022 IEEE 19th India Council International Conference (INDICON) · 2022
The availability of sensors in smart devices and the readiness to collect vital data have paved the way to applications that recognize human activities. The patterns formed by these data can be efficiently recognized by the use of deep learning networks. This study examines the two variants of recurrent neural network, namely, long short-term memory and gated recurrent units, in recognizing human activities using raw accelerometer sensor data as well as feature matrix computed from axis data. It is interesting that better accuracy at a faster rate can be obtained by training networks using feature vectors. The public datasets UCI-HAR and WISDM and data collected using smartphone application gave highest accuracy with the LSTM and GRU combination network.