Thrive Band

Inspiration

The inspiration for this project comes from the desire to help athletes and fitness enthusiasts optimize their training regimen and achieve peak performance, while also preventing injury and muscle cramps. We recognized the need for a more personalized and dynamic approach to pace calculation and progressive overload, as well as the importance of incorporating safety mechanisms to monitor the user's physiological state.

What it does

The proposed system consists of a novel approach to calculating an adaptive pace for runners and body lifters, allowing them to sustain their performance over different run lengths and overload progressively. The system also utilizes encoder-decoder LSTM neural networks to analyze sweat electrolyte readings and predict the likelihood of muscle cramps, providing users with valuable insights to help them break boundaries in their physical activity.

How we built it

We developed the adaptive pace calculation algorithm, which adjusts the baseline pace based on the new distance. For the muscle cramp prediction model, we leveraged the Ohio T1DM dataset to train an encoder-decoder LSTM neural network to predict future glucose levels from sweat electrolyte readings.

The physical device that houses the algorithms and components includes an Arduino mini, a water sensor, an LCD screen with buttons/dials, and a gyroscope/accelerometer. The schematics for the device are provided in the document.

Challenges we ran into

The main challenge was developing the predictive algorithm for muscle cramps based solely on sweat electrolyte readings, as this is a non-trivial task. We had to overcome the challenge of accurately predicting future glucose levels from sweat data, as the relationship between the two is not straightforward.

Accomplishments that we're proud of

We are proud of the novel approach to adaptive pace calculation, which demonstrates superior efficacy in maintaining consistent performance across diverse run lengths. We are also proud of the development of the muscle cramp prediction model, which leverages advanced machine learning techniques to provide valuable insights to athletes and fitness enthusiasts.

What we learned

Through this project, we learned about the importance of personalized and dynamic approaches to fitness and athletic training. We also gained insights into the challenges of predicting physiological states from non-invasive sensor data, and the potential of machine learning techniques to address these challenges.

What's next for Thrive Band

We have a variety of future implementations, such as more sensors (heart ECG sensor) to more accurately predict physical injuries. Another big improvement we plan to do is implement secure wifi capabilities to communicate wirelessly between the device and the web service, and to decrease the size of the device.

Built With

• encoder-decoder-lstm-nn
• next.js
• python