Action recognition of dance video learning based on embedded system and computer vision image

Extraction and unfettered online / offline video sequence to identify complex human activity is computer vision a challenging task. To presents the classification of Indian classical dance moves using the powerful features of embedded system tools: Field Programmable Gate Array (FPGA). In this work, the Indian classical dance video for human action recognition is, YouTube data from offline and online control audio and video recordings of live performances carried out. Handprint create offline data with ten different themes familiar dance of 200 m / from various Indian classical dance forms in the context of a variety of poses. Online data collection dance ten different subjects from YouTube. Each dance posture is occupied 60 or video in both cases. FPGA training and 8 different sample dimensions, each performed by a plurality of sets of subject. The remaining two samples for testing the trained FPGA. Different FPGA architecture design, and with our test data in order to obtain better recognition accuracy. Compared the report on the same data set and other classification model to achieve a 90% recognition rate.

Automated summary

The research focuses on classifying and recognizing Indian classical dance videos using FPGA, with data collected from offline manual creation and online YouTube sources. Training and testing were conducted using different subjects and poses, resulting in a 90% recognition rate compared to other classification models.