A Pilot Study on Continuous Breaststroke Phase Recognition with Fast Training Based on Lower-Limb Inertial Signals

Area of Science:

• Biomechanics
• Sports Science
• Wearable Technology

Background:

• Human underwater motion recognition traditionally requires manual configuration, which is time-consuming.
• Accurate segmentation of swimming strokes into distinct phases is crucial for performance analysis and assistance.

Purpose of the Study:

• To develop an automated method for continuous stroke phase recognition using lower-limb inertial signals.
• To reduce the time and manual effort involved in human underwater motion recognition.
• To enable real-time classification of swimming phases for potential aquatic locomotion assistance.

Main Methods:

• Proposed a continuous stroke phase recognition method utilizing lower-limb inertial signals.
• Developed a wearable sensing system with inertial measurement units (IMUs).
• Employed the K-nearest neighbor (k-NN) algorithm for classifying segmented stroke data (propulsion, glide, recovery).

Main Results:

• Achieved high accuracy with an average precision of 93.7% and recall of 92.6% across elite swimmers.
• Demonstrated feasibility with minimal training data (5 stroke cycles).
• Showed a low average time difference of 66.2 ms (4.2% of stroke phase) for key stroke event recognition.

Conclusions:

• The proposed method effectively automates continuous stroke phase recognition in swimming.
• The system shows significant potential for human aquatic locomotion assistance applications.
• The findings support the feasibility of using lower-limb IMUs for detailed swimming analysis.