Challenge 1

Most of the work in this area uses a combinations of sensors at multiple locations (ankle, chest, wrist etc). Most of these sensors are not available to users in real-time making such work impractical.
Therefore, we are targeting human activity detection using only a wrist watch which is very common nowadays (According to Pew Research Center, About one-in-five Americans use a smart watch or fitness tracker).
Combination of multiple sensors provides a good set of distinguishing features which makes it easy for the model to detect multiple activities.
Whereas, in this project, the limited set of sensors made it difficult for our model to distinguish between activities.
To fix this problem, we have used multiple approaches such as merging similar activities, adding untargeted activities, adding features like mean and standard deviation etc.

Challenge 2

We started our project using the MotionSense HRV wearable and the mCerebrum app to collect real-time data.
However, the app was giving a lot of problems. For example, we were seeing multiple data values for same timestamps in the csv file as shown in the following figure:

Model training and real-time inference are performed on data collected from different sensors.
To prevent the trained model from getting biased to data from one type of sensor, we trained the model on two different datasets (using different sensors).
We observed that using this combination of datasets improved our real-time inference accuracy significantly.

To train our model with good accuracy, we needed datasets with a large number of users and with our target sensors and activities.
Looking for such relevant datasets took a significant amount of time.