Fall detection app: Interpreting data
How do the fall and other activities compare?
In the previous post, I described how I obtained more than 3000 samples of different activities of volunteers walking, sitting down, laying down and falling on the ground with help of the phone sensors.
Measurements with adequate types can be previewed and interpreted with a small effort. To make it even easier, we can calculate the magnitude of the acceleration at any given time. We will apply this transform on all the timestamps to get only a 1D signal.
x,y, and z are axes of the acceleration data
Here are the charts 📉
Interpretation 📊
Walking 🚶
Walking alone can be described by repetitive peaks of moving weight from one foot to another, where a phone is in the pocket of the volunteer’s pants. The acceleration increase is the consequence of a foot hitting the ground. The acceleration decrease is the consequence of moving the foot up into the air to move it. The good news about walking is that it does not contain high peaks of acceleration as in comparison with the fall. The visible repetitive pattern is used to count steps on your smartphone too.
Laying down 🛏️
Laying down on the bed is represented by a little dip in acceleration, which is the act of the moving person downwards to lie down. The central part of the body hits the bed and results in a high peak. Thus, the sequence of other little peaks follows and is related to moving the legs on the bed, placing hands around, etc.
Sitting down 🪑
I found an interesting difference between sitting on a soft sofa and a wooden kitchen chair. Sitting on the soft couch could be understood as a dampening harmonic wave. Firstly, there is an acceleration descent when the volunteer starts the act of the movement. Secondly, the bottom hits the sofa and results in a slow incline of acceleration. After sitting down, the volunteer’s body springs up and down for a certain period in dependence on the sofa.
Sitting down on a hard surface chair is comparable to falling on the ground. There is no suspension, which would gradually transform the force of the moving person. The act of sitting is guided by a slight decline in acceleration and followed by a high peak.
Falling 🆘
As it was said, falling on the ground is comparable to sitting on a chair. However, there are many differences, which we will be taking advantage of in the future.
- The fall is represented by a higher first peak usually.
- The more prolonged decline of the acceleration before hitting the ground
- After the first significant peak, more subtle but still significant peaks follow (Other parts of the body are hitting the ground. E.g. hands, head or legs)
Conclusion
Different activities produce different sequences of acceleration. Here is my illustration of how we could generalize the activities.
In the next part, we will explore defining features and use them in the data.
