# Closing Comments

Now we have completed this project. This includes data collection, data labelling, label analysis, model building, model evaluation and edge code generation.

For this project, there are multiple paths to go further, such as, to deploy it on edge, try different configurations, to try to improve model performance or trying adding or changing gestures.

For deploying it on the edge, we might would like to reduce the model size while maintaining good accuracy. Then we can reduce the window size since the smaller the window size the smaller the first layer size of the model, but make sure the window is big enough to fit our longest event in it. The model can be used for a wide variety of applications such as control inputs for a wireless headset.

As mentioned in Data Collection, we used the configuration file found in the Capture Server repository for the radar to perform data collection. This configuration file can be further optimised. Please note that the newly collected data with the new configuration file will not be compatible with the old data. This is because by changing the configuration you are changing the sensor behaviour and the output will be different for the same observer phenomena. For more information regarding how to modify the configuration file, please refer to Acconeer's Sparse Service Documentation (opens new window). You can also use other services which each has different properties and works better for different applications.

For improving the model performance, one thing that we can do is to collect more data since the data in this project was mainly collected by one of our engineers, which means the model might be overfit on it. Or if we want to add different gestures on the top of this project, we can begin with collecting all the gestures that we wish to evaluate into one file and visualise the pre-processed data for gesture selection. We can also try to improve the preprocessing such as scaling the frequency spectrum, such that the model only learns the dominant part.

Have fun!