A Hybrid Deep Learning Framework using CNN and GRU-based RNN for Recognition of Pairwise Similar Activities
Md Sadman Siraj, Md Atiqur Rahman Ahad
Abstract
A challenging task in human activity recognition is to classify very naturally similar activities. In this paper, we introduce a unified deep learning model working as a hybrid framework of Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) modules to solve the problem of recognizing activities that are similar and which occur in pairs in their distributions. We have trained and tested our model on two datasets comprising of pairwise similar activities. The proposed framework has been successful as it outperformed most of the state-of-the-art models for this task. This hybrid model achieves activity recognition accuracies of 89.14% and 87.76% on the two datasets respectively and proves itself accurate and scalable. Contribution- A CNN-GRU deep learning framework for recognition of pairwise similar activities using multimodal wearable data.