Badminton Stroke Recognition Based on Body Sensor Networks

Abstract

A badminton training system based on body sensor networks has been proposed. The system may recognize different badminton strokes of badminton players. A two-layer hidden Markov model (HMM) classification algorithm is proposed to recognize 14 types of badminton strokes. In the first layer, we use acceleration magnitude of the right wrist to determine a threshold to detect strokes, and then, the HMM is applied to filter out nonstroke motions. In the second layer, we adopt the HMM to classify all the strokes into 14 categories. Experimental results show that the two-layer HMM can achieve good recognition accuracy. The effectiveness and feasibility of the two-layer HMM classification algorithm have been verified in a comparison. Index Terms-Badminton stroke recognition, body sensor networks (BSN), hidden Markov model (HMM).

Robust human activity recognition using single accelerometer via wavelet energy spectrum features and ensemble feature selection

“…x, y, z acceleration variance (Wang, Guo, et al, 2016) 13-15 6 Signal magnitude vector (SMV) (Cheng & Jhan, 2013) 16 7…”

“…The application of BSN also covers sports training. For example, a badminton training system which consists of left wrist sensor, right wrist sensor, waist sensor and right ankle sensor is designed in (Wang, Guo, & Zhao, 2016). The proposed system fused the data of four sensors at the feature level and utilized the double-layer hidden markov model (HMM) classification algorithm to recognize 14 types of badminton batting activities.…”

Body sensor networks for monitoring performances in sports: A brief overview and some new thoughts

“…In Ref., [33] various badminton strokes are recognized based on BSN. The system contains wireless inertial sensor nodes, wireless receiving node and an operating system (PC).…”

“…for horse riding (Equestrian sports) Quaternion update based on gradient descent algorithm; Equestrian motion tracking based on quaternion; Analysis of riding poster based on Equestrian [33] Different strokes (serve, clear, push, Sample set generation by wavelet denoising, badminton stroke detection, window lob, chop, rushing, hook for fore-segmentation, feature extraction and selection; classification by a two layer HMM classifier hand & backhand) in badminton (Layer 1: verify if the activity is a stroke, Layer 2: multiclass classification of the strokes)…”

“…From Table 7, it is found that all studies have not addressed the computational complexity even the detailed papers with no other problems have missed only the computational complexity (e.g. in [33] ). This is an important issue to take into account for real-time implementation of a BSN system besides its performance.…”