Category: Uncategorized

This week, my team gave final presentations, and we all made video clips for the 10 minute demo video.

In my video clips, I filmed the procedure of data collection. Dropping the dummy forward, backward, and collecting the data via walking.

This week, I collected data to improve the variety of our fall data set. Collecting a better variety of data can improve fall detection because SVM’s make decisions on margin vectors. Collecting more varieties of falls can have a significant impact on margin vectors. As well, I started work on the slides, demo video, and final paper with the rest of the group.

This week, I worked on integration of the SVM code with the Raspberry Pi with Sojeong and Max. I made sure to discuss with Max the expected input and output of the SVM code so it would be runnable on the Raspberry Pi, and then Sojeong and I made the necessary changes for Max.

As well, I began plotting dummy falls and human falls so I could see if there was any significant visual difference between the falls. From visual inspection, I could see no observable differences.

This week I worked with Jacob and Soejong on porting/integrating the SVM to the RPi and separately worked with Nick on getting a mobile Bluetooth app online.

The SVM is running nicely on the phone and seems to be working as expected with the IMU data — I have it reporting results in realtime over Bluetooth to the generic Bluetooth serial terminal app from before. Before running in realtime, I determined that each inference takes about 13ms which is well within acceptable bounds — right now, we’re checking a 1s window once a second. Determining power draw/battery life probably won’t happen since our battery has disappeared (somewhere in mail system, not on route to me), but I might be able to work something out.

On the app front, I was able to run an older, native version of Nick’s app, so he has a working environment. That uses a native library, so the biggest hurdle (getting arbitrary native code working) is done.

This week (and weekend) I spent far too long trying to get a simple Bluetooth app working. Some combination of my unfamiliarity with the Android ecosystem, Java, Javascript, JSX, and the toolchains used to work with all of those led to a lot of time and no real progress. I asked Nick to take a look at it, but the environment he’s been using (Expo) doesn’t seem to be compatible with the sort of device-native code needed by the various Bluetooth approaches I’ve tried.

In order to make real progress, I’ll ignore that for now (other than working with Nick as needed) and get to running the SVM on device — that should be much smoother.

This week, our group demoed our current progress.

After the demo, I started work on integration of the machine learning algorithm with the Raspberry Pi, Max, Sojeong, and I started discussing how max would like the inputs and outputs of our code to be presented, so that when he runs our code on the Pi, his code can interface with our code.

This week, I collected data with the dummy. I collected 30 minutes of walking data, 30 minutes of falling data, and 30 minutes of sitting data. I then had to label the data. As well, I worked on plotting frequency features of falls. I noticed the frequency wavelet plots of fall segments had a specific appearance indicating these features will be beneficial.

This week I brought up a fairly robust simple two-way messaging system between an Android phone and the device. This has error handling and retry logic to make sure that it doesn’t get confused/softlocked due to a flaky connection — having a means to power cycle it without pulling the plug or sshing in is nice.

This was all done on the device’s end; I used an existing app for the Bluetooth serial handling on the phone. This upcoming week, I’ll work on a basic React Native app to work as an example receiving end of the interface in lieu of directly integrating with Nick’s work.

In parallel, I’ll work to get the fall detection SVM running on the RPi. this shouldn’t take much effort, but it may reveal performance issues. If all goes well, I should be able to get the realtime detection working on the IMU data and/or record data to test.

Other than working with the team on refocusing, I didn’t get much done this week — the RPi and IMU showed up in the mail from Nick on Friday.

This week I’ll be figuring out & writing a simple app and the RPi Python code to talk to it over Bluetooth. This will constitute the interface between the hardware subsystem and the app/ML.

Once I have that data connection working, I’ll start collecting data to send to Sojeong and Jacob.

This week, The test dummy arrived to my home. I commenced collecting fall data with the test dummy.

As well, I designed a system to adjust the parameters of the FFT and Wavelet frequency features which will be fed into the SVM, and also made a system to adjust the parameters of PCA compression that will happen during pre processing of features.