Currently, in scientific journals ranging from sports medicine to kinesiology, much research has been conducted to study the EMG data of archers. When our team joined up for MHacks, we had plans to test out the cool Myo Armband that one of us owned. As we brainstormed different projects we could pursue with this technology, we were surprised to find that two of our four-member team were involved in archery. One teammate competed all through high school, and had also coached the sport. One of the most challenging aspects of coaching, he told the rest of us, is being able to spot a faulty posture. Archery is a precision sport, and the slightest errors in form -- a lowered elbow, a small push during release, an unnoticeable tremor during the draw -- will significantly alter the accuracy of the shot. Even the most skilled coaches may not be able to catch these minor slips.
With all of these coincidences in mind, we were very excited to hack the Myo band. We decided to create an automated coach for novice archers. The band, which is very sensitive to the muscles in the forearm, will better detect an archer's small errors than will a coach's eyes. The infrastructure of our hack goes as follows: a Myo Armband worn by an archer sends EMG data to a Java application, which analyzes and sends signals through sound clips and vibrations. When the archer is in the correct posture, he or she will receive positive feedback through short vibrations in the Myo. If the archer is taking a long time to release, the application will also play a blip of sound every 1.5 seconds until release of the bow string, ensuring that the archer's timing is also accurate.
One of the most challenging aspects of the hack was parsing through the EMG data. The analysis had to parse through the "noise" and find most important data points -- namely, when the archer reached the correct posture. We needed to design a program that could analyze the data and consistently recognize the different positions that an archer took. Our algorithm took the averages of the readings from the eight sensors to recognize when an archer was in "ready" position, whether or not his or her elbow was too high, and whether or not the release was well-executed. We went through several trials with a Nerf bow to set the thresholds between a "good" grip and a "bad" one, etc. (We used a Nerf bow because, even though we had access to a real bow, safety measures and precautions had to take precedence.)
If we continue this project, we wish to extend the application to one that can be used on a mobile device, as this feature will be more accessible to an archer on a range. Unfortunately, given that the raw EMG data is not available for mobile devices, we were unable to achieve this during the hackathon. However, we were nevertheless able to create a working prototype in only 36 hours. We were able to take advantage of the versatility of the Myo band and build a real-world application.