This paper aims to design an accurate and robust algorithm for counting the steps based on the smartphone's accelerometer for indoor applications. Different daily activities have been considered in the experimental scenarios, including normal walking with various smartphone positions and running activities. The detection process's accuracy is ensured by setting an appropriate segregation approach and defining a fixed threshold for each experiment. The algorithm generates a new envelope signal that mimics the collected steps signal without the vibrations and noise elements to avoid miscounting the exact steps. The followed approach results have been compared with the pedometer applications on the android platform and HUAWEI Watch GT-731. The proposed algorithm showed improved detection accuracy in contrast to the other literature approaches. The obtained accuracy was around 99% for the standard walking scenario and approximately 97% for the running activity scenario.