Author: carysd

This week, I worked to integrate feedback from our Design Presentation and Weekly Lab Check-in, specifically regarding our calibration process. Rohaib and I iterated our calibration process so it still includes the user setting their initial baseline, both for their upper and lower back posture orientation, as well as their feedback threshold for position/angular change over a specified time, but now also includes the desktop app guiding the user through a series of movements to confirm this threshold. The idea is that having users turn left, right, and then stretch back allows them to complete “basic” or “everyday” movements and determine whether their current threshold settings will notify them of worsened posture, and whether they would like that to happen.

Additionally, as a team, we met with the CMU PhD student in HCI I contacted, who has completed research on posture modelling. He recommended having the user define their “good” vs “bad” posture, which we were already planning to do, and suggested allowing the user to adjust their feedback threshold throughout PosturePal use. We had not originally considered this idea; however, we agree that implementing it would allow users to better specify their haptic feedback throughout use, building on the desktop-app-led initial movements to trial the user’s set threshold.

I believe I am on schedule. I am hoping the vests we ordered last week will be delivered over Fall Break (as unfortunately they were not delivered this week), so I can begin editing and reassembling them once I am back. Especially since this week Allison was able to trial connecting two IMUs to the MCU and demo angle changes with cardboard, triggering haptic feedback. As a result, after Fall Break, the next step is to build the vest, trial-attach the IMUs to it, and see how PosturePal handles actual spinal changes.

After completing the design presentation this week, I worked to implement and think about the questions and feedback I had been given at the end of our presentation. Specifically, the question about our current vest design and how we will properly and securely connect all of our components to it. As a team, we were not worried about attaching all components to the vest, however, ensuring the IMUs are correctly aligned with each vertebra and adjustable for every user is. As a result, I spent a large amount of my time this week refining our vest design, specifically how we will edit the vest to be adjustable and attach relevant components to it. Our working design now includes the use and reconstruction of two vests, in which we will use their original mesh components to enclose and secure our IMUs and PCBs, as well as use the various adjustable straps of the two vests to make the location of each IMU flexible for users. I then ordered our vests, and our updated parts lists and order status of each item can be found below.

Once our first round of parts was delivered this week, Allison and I began assembling and testing the pieces we had received, for example, soldering the MCU. Once the vest has been delivered, my goal for next week is to begin editing it so that the assembled parts can be easily attached and removed for iteration. Specifically, editing the vest to allow for adjustable IMU integration and PCB placement. As of right now, I feel like my work is on schedule; however, I am worried about my progress for next week. Once everything has been delivered, I would like to make good headway on my work so I feel confident about what I have completed for the project as I come back from Fall Break. Finally, I received a response from a CMU professor I contacted about their past research in posture monitoring, and they connected me with a CMU PhD researcher on the topic, who my team and I have a meeting with next week.

This week I worked with my team to finalize our designs for the proposal presentation, including where we will locate the IMUs (now on the T3, T12, and L5) as well as potentially moving our haptic motors for feedback, from their original location of next to each IMU, to now being placed along the shoulders of the vest (to prevent added noise to the data collected from these IMUs when these motors shake to get the user’s attention). Additionally, I looked into the VQF library we plan to use, including reading the associated research paper, to better understand its potential for integration into our design and the values it would require from our IMUs to be successful. As we discussed in our weekly check-in, the filtering system takes in values from both a gyroscope and an accelerometer, and works to prevent gyroscope bias by calculating a small offset from the gyroscope values and subtracting that from the readings.

Moreover, as recommended in our check-in this past week, I looked into potential experts/people with experience in the posture field to contact for advice or opinions on our current approach and user problem. After talking with my team, I contacted three researchers in the space connected to CMU who published the paper “Robust, Low-cost, Non-intrusive Sensing and Recognition of Seated Postures” and we are hoping to meet with them in the next two weeks. Right now, I feel on track with my schedule, but I am anticipating my workload to increase in the next couple of weeks once we have all of our parts in for PosturePal, and I can start attaching and testing the IMUs. In the next week, I hope to secure meeting times with each of the researchers we contacted, as well as wire the IMUs (if delivered).

My work this week centred on further research into our IMU placement, data collection, and calculations for proper integration into PosturePal. We are considering using LSM6DSOX as our IMUs due to their six-axis capability, low power consumption (0.55 mA in their higher-performance mode), and compact size (2.5 mm x 3 mm x 0.83 mm). As Allison mentioned in her weekly report, one potential drawback of these IMUs is that they do not allow for wireless communication, and as a result will integrate either I²C or SPI into our current design. As done in a 2021 research article by Paloschi et al, we will be placing IMUs on the T3, T12, and S1 vertebrae, which in the paper were compared to and validated against a motion capture system to detect poor spinal posture. The T3 and T12 movement and difference were used to capture thoracic changes, while the T12 and S1 were used for the lumbar. However, the research of  Paloschi et al did not use a wearable vest for participants when capturing this data and instead placed the IMU’s directly on the skin. As a result, we must and are working to take into account this difference in our design, specifically, how we will test for accurate measurements and worsened posture haptic feedback. One way we are considering to combat this in our design is that instead of integrating the three-axis acceleration and angular velocity given by the LSM6DSOX to calculate the IMU position over time (comparing each sensor’s new position to the location of its old one), we will compare the relative location of each sensor to the others. This way, we can determine if the upper back has curved in comparison to the middle of the spine, and similarly for the lower back. These relative calculations will also guide our calibration phase of PosturePal to identify what the user believes is their “ideal posture”. Specifically, we believe this will help us better meet our requirement of accurately identifying 90% of actual slouches or occurrences of posture worsening over stretching or other movements. 

As a result, in our current design process, we believe we are on time and in a good position to complete our design proposal with an updated approach to sensor data calculations and approach. In the next week, my goal is to finalize the positions of these IMUs and the communication protocol they will be using with our MCU and software stack with my teammates to ensure the data we are collecting and the calculations we are performing will best allow for accurate haptic feedback.