Human activity recognition (HAR) applications have received much attention due to their necessary implementations in various domains, including industry 5.0 applications such as smart homes, e-health, and various Internet of things (IoT) applications. Deep learning techniques have shown impressive performance in different classification tasks, including HAR. Accordingly, in this paper, we develop a comprehensive HAR system based on a novel deep learning architecture called Multi-ResAtt (Multilevel residual network with attention). This model incorporates initial blocks and residual modules aligned in parallel. Multi-ResAtt learns data representations on the Inertial Measurement Units (IMUs) level. Multi-ResAtt integrates a recurrent neural network (RNN) with attention to extract time-series features and perform activity recognition. We consider complex human activities collected from wearable sensors to evaluate the Multi-ResAtt using three public datasets, Opportunity, UniMiB-SHAR, and PAMAP2. Additionally, we compare the proposed Multi-ResAtt to several deep learning models and existing HAR systems, and it achieved significant performance.