2-channel parallel face detection board using two TMS320C6201 DSPs

Author

Chen, Bo-Ya ; Su, Guang-Da ; Deng, Ya-Feng

Author_Institution

Dept. of Electron. Eng., Tsinghua Univ., Beijing, China

Volume

7

fYear

2005

fDate

18-21 Aug. 2005

Firstpage

4446

Abstract

This paper describes a 2-channel face detection board which can detect human faces in real-time video stream with high accuracy. Based on Viola et al.´s work, we mainly use a cascade-structured classifier to evaluate candidate windows, with a face template matching before to reject the obvious non-face candidates for accelerating. This detection arithmetic is implemented on hardware with two TMS320C6201 digital signal processors (DSP), which execute their detection tasks in parallel. The peak processing speed can get up to 24 images per second of 384×288 pixels for each channel. The false detection rate is below 1%.

Keywords

digital signal processing chips; face recognition; image matching; parallel processing; real-time systems; video streaming; 2-channel parallel face detection board; AdaBoost algorithm; TMS320C6201 DSP; candidate window evaluation; cascade-structured classifier; detection arithmetic; digital signal processors; face template matching; parallel image processing; real-time video stream; Cameras; Digital signal processing; Face detection; Face recognition; Hardware; Humans; Machine learning; Signal processing algorithms; Streaming media; Support vector machines; AdaBoost algorithm; DSP TMS320C6x; Face detection; Parallel image processing;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning and Cybernetics, 2005. Proceedings of 2005 International Conference on

Conference_Location

Guangzhou, China

Print_ISBN

0-7803-9091-1

Type

conf

DOI

10.1109/ICMLC.2005.1527722

Filename

1527722