Posted on

Nia’s Status Report 3/28

This week, I updated the frontend to not receive feedback from the user about whether or not they signed correctly when the model identifies the sign as incorrect. This is to prevent our model from possibly receiving incorrect information and as a result, decreasing the accuracy of the model.

I also began writing the data collections script that we will use to semi-automate our data collection process. Specifically, this script will allow us to automatically save .csv files with proper naming mechanisms as well as pause data collection at different points and resume from where we left off (for example, complete 20 trials of “A”, end the terminal process, start the terminal process, continue at trial 21 of “A”.

I addition, I helped brainstorm how we incorporate the touch sensors on the glove such as positioning and wiring.

Next week, I plan on completing the data collection script so that once we have the hardware and processing complete, we will begin collecting data.

Posted on

Teadora’s Status Report for 3/28

This week I assembled the glove by attaching the flex sensors and imu along with some cables. I also helped Kat with testing and calibration by wearing the glove. As I mentioned last week, we’re looking into using touch sensors to help identify some signs better. When I was deciding what parts to order, I chose the capacitive touch sensor breakout instead of pressure touch sensors like (https://www.adafruit.com/product/166).  The touch sensor breakout is built as an i2c board, which means it can integrate quickly with our existing setup. The pressure sensors would have needed  an ADC board and similar setup to the flex sensors, since both are variable resistors. We have enough funds, pins, and physical space to use an ADS, but I still chose the capacitive touch sensor because it seemed easier to directly set it up as binary touch/no touch readings, which is what I understood the design required instead of variable pressure. I also found that the size and shape of the easily available pressure sensors didn’t seem like it would work well with the glove — the circular pads were larger than a finger tip and seemed to get in the way, and the square ones were too small to work on the palm. With the capacitive touch sensors, I should be able to  tape out conductive regions with copper foil tape and detect when the finger makes contact with those regions. The foil tape arrived yesterday (Friday 3/27) so I’ll use that in testing this week. I feel like we’re on track with the project, assuming we can start to record data soon.

Posted on

Kat’s Status Report 3/28

This week, I worked on processing the ADS and the touch sensors as well as the calibration of the readings coming from the sensors mounted directly to the glove. Now we have a complete set set of readable data from the sensors. I also started work on the wireless transmission of the RBi although I do not think it will be ready by the interim demo. On the hardware side I soldered all the header pins onto our boards like the Raspberry Pi and the touch sensors board so they can be mountable onto the mini breadboard.

This week we have the interim demo so I will work on what I will talk about during it. Nia and I will continue working on the machine learning model as we are thinking of shifting from a neural network to a KNN model.

Posted on

Team Status Report 3/28

The current risks are related to the time it takes to train the ML model. Touch sensors won’t affect the model since it will just be added as an additional feature so they will likely improve it and the ml model shouldn’t take more than a week. Training should take max 5 mins once we have data, so the main holdup really is getting data. We can improve the model post MVP by tuning training parameters

This past week we assembled a glove to make sure the flex sensor readings worked on the hand.

We added capacitive touch sensors to the system so we could better detect when fingers are touching each other or the palm, which is necessary to distinguish signs like U/V, M/N, etc. Teadora selected capacitive touch sensors instead of force sensing resistors and details the reasons for the decision in this week’s status report. There wasn’t a significant parts cost ($15 at most for conductive tape and a breakout board), but we will need to test the touch sensors and make sure the readings can transfer easily to the ML model. We will display our in progress glove with flex sensors and IMU at the interim demo.

Posted on

Kat’s Status Report 3/21

I finished the reading of the data from the flex sensors and IMU. I processed that data into percentages of finger bends with a pretty accurate range. For our project we need at least 2 classifications of bend which I have achieved through calculations right as we receive data. I also made a calibration system for the glove that we can use if needed for different users. Nia and I are also working on the classification of different signs based on binary vectors we created. Teadora and I attached the RPi to a breadboard and can hook up the gloves with its attached flex sensors to it for testing.

This week I will finish the ADS connections so we can read all of the analog flex sensor values. The IMU’s data is a little messed up, I need to finish calculating the position values as they are not correct as of right now. However, it does process motion well and in the correct direction. Nia and I will be working on data collection for our model now that we have started glove assembly.

Posted on

Team Status Report 3/21

Now that we know the sensors can receive data, we’re going to move quickly towards assembling the glove so we can train the ML model. We’d like to have it assembled soon so we can have something ready for the interim demo in a little over a week. We’re considering adding capacitive touch sensors to the glove (which we’ve already ordered for testing) to make sure we can accurately represent all of the signs.

The most significant risk that could jeopardize the product is how well the sensors  hold up during testing and data collection. We will prioritize keeping the sensors in place and secure during data collection to ensure we collect accurate data across all trials. If the sensors move significantly during data collection, we run the risk of our model incorrectly assigning classes.

Posted on

Nia’s Status Report 3/21

This week, I attended the ethics lecture and discussed with my team along with others how we can be mindful of ethical tradeoffs when creating new technology for society. Along with that, I confirmed the set up of our feature vectors and the machine learning pipeline we plan to implement. Next week, we will assemble the glove and I will test the model’s evaluation API that connects to our frontend interface.

Posted on

Teadora’s Status Report 3/21

This week I attended the ethics lecture and confirmed backup parts orders. Last week I was concerned about how the voltage divider on the flex sensors would interact with the input impedance to the ADC, so I tested a TL072 op amp as buffer. It turned out this wasn’t necessary after I worked with Katherine to get the flex sensor calibration working on the build in Pico ADC pins. We also connected IMU to Pico with I2C and were able to read out data. Confirmed that the connections worked for battery power and microusb power. My part in this was measuring voltage, connecting the test setup, and referencing suggested designs in spec sheets. Tomorrow (Sunday 3/22) we’re going to assemble the glove and try to get preliminary data readings.

https://photos.app.goo.gl/PmM3M6UYRMjM3gnv5

https://photos.app.goo.gl/sXXgsgqtNRei6V7HA

Posted on

Team’s Status Report 3/14

This week, Kat did the processing for the Raspberry Pi Pico to receive analog information and send it to the machine learning model. Nia generated a synthetic dataset that she ran the neural network with. Teadora tested the flex sensors and will be continuing to evaluate them and the IMU with the Raspberry Pi Pico this upcoming week.

This week we will make sure we can receive the analog signals properly from the IMU and flex sensors. We will also continue working on the machine learning model and optimize it for the types of data we will be using. We also will complete our work on the data organization such as how the data goes from when it first arrives to on the computer to the machine learning model.

Posted on

Katherine’s Status Report 3/14

This week I finished the wired connection between the Raspberry Pi Pico so we can read values from the analog signals. Once the ADC adapter comes in we will be able to connect all of the analog signals at once. I plan on finishing the bluetooth processing tomorrow so we can focus on the machine learning model for the rest of the week and time.