This paper proposed smartphone-based human activity recognition (HAR) using deep learning (DL). Smartphones offer numerous benefits, including enhancing health monitoring, providing personalized fitness recommendations, improving safety for the elderly or at-risk individuals, and managing workplaces efficiently. By analyzing movement patterns, these Smartphones apps can detect anomalies, encourage active lifestyles, and offer valuable insights for health and lifestyle management. Traditional models using smartphone accelerometers for HAR struggle with multimodal, high-dimensional sensor data in complex scenarios. DL is a promising solution to many recognition problems, including speech, image, and natural language processing. Its unique feature is that it automatically learns complex features from huge data sets, allowing for more accurate and effective recognition than using traditional machine learning. Therefore, this paper applied DL to enhance the HAR problem. This paper used UCI’s HAR dataset which uses an architecture based on accelerometer data from smartphones. The research focuses on identifying six common daily activities, including standing, walking, sitting, running, and downstairs. Thirty volunteers, ages 19 to 48, participated in the experiments by executing the six different activities. Data collected from smartphones and wearable sensors helps in accurately recognizing and differentiating these activities. In this paper, multiple machine learning and DL models are deployed. Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) and CNN and LSTM models showed higher accuracy than other models. Moreover, the Deep Q-Learning model has achieved promising results and proven its effectiveness compared to traditional models in recognizing the HAR problem.