Duo Streamers: A Streaming Gesture Recognition Framework

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhu, Boxuan, Yang, Sicheng, Wang, Zhuo, Liang, Haining, Shen, Junxiao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917935826599936
author Zhu, Boxuan
Yang, Sicheng
Wang, Zhuo
Liang, Haining
Shen, Junxiao
author_facet Zhu, Boxuan
Yang, Sicheng
Wang, Zhuo
Liang, Haining
Shen, Junxiao
contents Gesture recognition in resource-constrained scenarios faces significant challenges in achieving high accuracy and low latency. The streaming gesture recognition framework, Duo Streamers, proposed in this paper, addresses these challenges through a three-stage sparse recognition mechanism, an RNN-lite model with an external hidden state, and specialized training and post-processing pipelines, thereby making innovative progress in real-time performance and lightweight design. Experimental results show that Duo Streamers matches mainstream methods in accuracy metrics, while reducing the real-time factor by approximately 92.3%, i.e., delivering a nearly 13-fold speedup. In addition, the framework shrinks parameter counts to 1/38 (idle state) and 1/9 (busy state) compared to mainstream models. In summary, Duo Streamers not only offers an efficient and practical solution for streaming gesture recognition in resource-constrained devices but also lays a solid foundation for extended applications in multimodal and diverse scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2502_12297
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Duo Streamers: A Streaming Gesture Recognition Framework
Zhu, Boxuan
Yang, Sicheng
Wang, Zhuo
Liang, Haining
Shen, Junxiao
Computer Vision and Pattern Recognition
Gesture recognition in resource-constrained scenarios faces significant challenges in achieving high accuracy and low latency. The streaming gesture recognition framework, Duo Streamers, proposed in this paper, addresses these challenges through a three-stage sparse recognition mechanism, an RNN-lite model with an external hidden state, and specialized training and post-processing pipelines, thereby making innovative progress in real-time performance and lightweight design. Experimental results show that Duo Streamers matches mainstream methods in accuracy metrics, while reducing the real-time factor by approximately 92.3%, i.e., delivering a nearly 13-fold speedup. In addition, the framework shrinks parameter counts to 1/38 (idle state) and 1/9 (busy state) compared to mainstream models. In summary, Duo Streamers not only offers an efficient and practical solution for streaming gesture recognition in resource-constrained devices but also lays a solid foundation for extended applications in multimodal and diverse scenarios.
title Duo Streamers: A Streaming Gesture Recognition Framework
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2502.12297