Pengembangan Pengenalan Aktivitas Manusia Secara Real Time Menggunakan Metode Convolutional Neural Network dan Deep Gated Recurrent Unit

Feren Ade Verilia
Rangga Firdaus
Hery Dian Septama

Ulil Albab
Vol. 2 No. 2: Januari 2023

Recurrent neural network (RNN) have achieved great success in processing sequential data and yielded the state-of-the-art results in speech recognition, digital signal processing, video processing, and text data analysis. In this thesis, proposed a human action recognition method by processing the video data using convolutional neural network (CNN) and deep gated recurrent unit (GRU) network. First, features are extracted from frame every multiple of six in the videos to helps reduce the redundancy and complexity. Next, the sequential information among frame features is learnt using deep GRU network, where multiple layers are stacked together to increase its depth. The result of this study achieved 92.01% F1 Score on YouTube 11 Actions dataset. This method achieve 65.43 FPS.