Continuous Human Activity Recognition With Distributed Radar Sensor Networks and CNN–RNN Architectures

Unconstrained human activities recognition with a radar network is considered. A hybrid classifier combining both CNNs and RNNs for spatial-temporal pattern extraction proposed. The two-dimensional (2D-CNNs) are first applied to the data perform spatial feature on input spectrograms. Subsequently, gated recurrent units bidirectional implementations used capture long- short-term temporal dependencies in maps generated by 2D-CNNs. Three NN-based fusion methods were explored compared utilize rich information provided different nodes. performance of proposed was validated rigorously using K-fold CV L1PO method. Unlike competitive research, dataset continuous seamless inter-activity transitions that can occur at any time unconstrained moving trajectories participants has been collected evaluation purposes. Classification accuracy about 90.8% achieved nine-class HAR halfway