The use of wireless sensors to monitor physical activity is an effective way to identify postureand motions in everyday situations. In this research, Deep learning models like BiLSTM_Dense, Dense, and LSTM_Dense have been used to give a simple and reliable classification of different physical activities. Sensors are used in our technique and are placed on the individuals’ wrist, chest, and ankle. From the signals captured by these sensors, twelve different physical activities were classified using comparative analysis of each model’s performance. The collected dataset is made up of recordings of ten participants participating in 12 different physical activities. Each volunteer has a unique profile. The classification results demonstrate strong validity, with accuracy (positive predictive value) and recall (sensitivity) over 90% for all physical activities..