This paper provides a comprehensive survey of deep learning-based human activity recognition (HAR) using sensor data. It highlights the recent advancements in three main areas: sensor modality, deep model, and application. The authors discuss the advantages of deep learning over traditional pattern recognition methods, which often rely on heuristic feature extraction and struggle with unsupervised and incremental learning tasks. They survey various deep learning models, including deep neural networks (DNN), convolutional neural networks (CNN), autoencoders, restricted Boltzmann machines (RBM), and recurrent neural networks (RNN), and their applications in HAR. The paper also addresses the challenges and future directions in deep learning-based HAR, such as online and mobile deployment, unsupervised learning, high-level activity recognition, lightweight models, non-invasive sensing, and beyond activity recognition applications. The authors conclude by emphasizing the potential of deep learning in advancing HAR and suggesting areas for future research.This paper provides a comprehensive survey of deep learning-based human activity recognition (HAR) using sensor data. It highlights the recent advancements in three main areas: sensor modality, deep model, and application. The authors discuss the advantages of deep learning over traditional pattern recognition methods, which often rely on heuristic feature extraction and struggle with unsupervised and incremental learning tasks. They survey various deep learning models, including deep neural networks (DNN), convolutional neural networks (CNN), autoencoders, restricted Boltzmann machines (RBM), and recurrent neural networks (RNN), and their applications in HAR. The paper also addresses the challenges and future directions in deep learning-based HAR, such as online and mobile deployment, unsupervised learning, high-level activity recognition, lightweight models, non-invasive sensing, and beyond activity recognition applications. The authors conclude by emphasizing the potential of deep learning in advancing HAR and suggesting areas for future research.