Smartphone based Step Detection( The popular approaches )

Smartphone step detection is usually achieved using the smartphone accelerometer data. The information is then filtered to reduce noise. Later using the filtered data some look for peek values others for zero crossings to identify steps. I tried using the peek detection. Following are the results i came up with.

The Smartphone Accelerometer provides accelerations along x, y and z axis. The coordinate planes in which they act are given below.

The readings along the three axis during a 10 step walk with a sampling rate of around 300 milliseconds are displayed in the charts below.

As displayed  above the readings show that a step include a high peak. And that the number of peaks are similar to the number of steps. But though visually it is possible to analyze the steps, during the actual implementation using a threshold value to detect peaks, the results were highly inaccurate. Because the accelerometer is highly sensitive, even the smallest jerk while holding device can result in reasonable reading. 

However apart from the drawbacks it is clear that steps can be detected by analyzing the patterns of the reading.