With the world’s aged population increasing rapidly, an earnest need for recognition of human activities is sought after by the Healthcare units across the world. Applications of Human Activity Recognition include surveillance, fitness among various others. This manuscript briefly discusses the approaches followed for recognising human activities in the literature. Additionally, in the present work a lightweight RNN (recurrent neural network) model having LSTM (long short term memory) cells is proposed for HAR on the dataset. This lightweight RNN model employs a feature extraction task. Feature extraction is done in three domains—Temporal, Spatial, and Spectral domains. Recently published dataset KU-HAR is used for the presented work and the proposed work performs fairly well over the dataset.