The elderly and disabled often live without direct supervision and rely on remote monitoring for their care. However, current methods for information processing and activity recognition in healthcare decision-making have limitations. This paper proposes a professional solution by introducing a Deep Convolutional Neural Network (DeepCNN) and Fuzzy Support Vector Machine (FSVM) to extract significant features and label activities using a Fuzzy Membership Function (FMF). By utilizing FSVM with FMF, the problem of assigning instances to the correct class is solved, resulting in more accurate classification and a flexible system.The proposed method surpasses previous approaches on DLAs and UCI datasets, achieving an accuracy of 97.22% for DeepCNN-FSVM and 96.81% for FSVM on the UCI dataset. This research emphasizes the importance of activity recognition in smart healthcare for the elderly. Recent efforts by researchers have focused on developing high-performance methods in activity recognition to overcome challenges faced by previous approaches. Overall, this paper highlights the potential of the DeepCNN-FSVM approach in improving activity recognition in smart healthcare for the elderly and disabled.