Residual Connection Based Deep Neural Network for Evaluation of Athlete Competitive Ability

The competitive performance capability is on-spot performance of athletes comprehensive utilization of tactical and technical, psychological and physical capabilities. Due to athletes competitive performance included various performance contents and factors of influencing, that is critical for comprehensive and objective utilization of one qualitative or quantitative statistics in measurement and evaluation. The research proposed the Deep Learning (DL) based athlete competitive performance capability method by performing a time series moving pressure measuring information. The method architecture is developed that directly analyzes actual data and generates evaluation outcomes that hugely facilitates practical usage. The multiple scale feature extraction method is assigned through learning features in various scales. The Residual connection Neural Network structure is developed through short-cut connection and Deep Neural Network (DNN) is much effectively trained. The evaluation of real athlete competitive performance capability tests is compared along various related techniques. The proposed method attained testing accuracy of 96.5% in Task 1, 97.5% in Task 2, 98.3% in Task 3 and 98.5% in Task 4 for 1200 testing samples.