Human activity recognition (HAR) is considered as one of the most difficult and challenging issues now a days. Many experiments are now in progress regarding this problem. Among many human activities, mostly six are considered for research in this area. This activity recognition issue can be measured with the help of smartphones and smartphone sensors, along with the connection of Internet of Things (IoT) devices. In this research, an improved deep learning scheme is proposed for the recognition of human activities. A customized Neural Network (NN) model was designed and tested for the research. The proposed model obtained 96.47% accuracy on the HAR with smartphones dataset that is better than most other analyzed models. Sensors such as accelerometer, gyroscope are focused on the data analysis portion of this research work. This article will give a clear idea of the dataset, Machine Learning algorithms, and the effect of the proposed algorithm.