Virtual throwing action recognition based on time series data mining in an augmented reality system

Introducing the similarity discovery algorithm based on time series into virtual throwing action recognition is a new attempt. Four key sub-patterns from the whole throwing action process were defined, and the corresponding sub-patterns discovery algorithms were designed. The similarity matching between the target time series and the template was done to implement the recognition of virtual throwing. The experiment results show that the proposed algorithm with less calculation, high recognition rate, strong anti-jamming ability, could meet the augmented reality system well in real-time.