Human activity recognition (HAR) has become one of the most essential topics of discussion nowadays. The main motto of HAR is to recognize several complex day-to-day human transitions and activities efficiently. It is basically a technology that analyses data from multiple sensors and visual devices to recognize distinct human motions and behaviors. The mechanism of HAR models comprises sensors and various intelligent algorithms that perform the job of human activity detection in an integrated way. HAR has gradually become an interesting area of research. Several research works have already been conducted in this domain, with many more in progress to increase the efficiency and accuracy of the model as well. HAR has numerous applications in domains such as health care, the Internet of Things, smart homes, security, entertainment, and human-robot interaction. This chapter focuses on several aspects of HAR by reviewing various literatures. Hence, this chapter provides new ideas that can deliberately contribute more in the domain of HAR utilizing several hybrid deep learning (DL) models. Moreover, this chapter also provides important highlights and findings through a detailed analytical comparative study. Possible future works that can be done in this domain are also stated here. We discuss the many types of sensors and data sources utilized in HAR, as well as the contribution of machine learning and DL approaches in HAR. We also discuss some of the open issues and future directions to enhance the accuracy, robustness, scalability, and interpretability of HAR systems.