Towards a feedback system for upper body bodyweight exercises using multiple inertial measurement units : A user-centred approach

Sammanfattning: This thesis explores the feasibility of developing an affordable and easy-to-use feed- back system that leverages information from multiple inertial measurement units (IMUs) to identify mistakes during upper body bodyweight training and provide real-time feedback to the user. To develop the system, a human-centered approach was used, which involved conducting semi-structured interviews with movement ex- perts and a workshop with targeted end-users to understand their needs. The study also involved 12 volunteers who performed push-ups, tricep dips, and planks cor- rectly and then made specific mistakes intentionally while wearing five IMUs. Binary and multiclass classifiers were developed to classify the exercise technique. The re- sults showed that personalized multiclass classifiers produced good to excellent quality results, while global classification techniques performed poorly. Additionally, an increased number of sensors did not always lead to improved classification re- sults, and the placement of the sensors could have a significant impact. The user in- put and evaluation of the feedback system pinpointed the need for customization and accessibility in the design. This thesis contributes to the development of a feedback system that can help users identify mistakes in their upper body bodyweight exer- cises and improve their form and technique.