BROWSE
Detail
Conference Paper
조명에 강한 실시간 얼굴 검출 ASIC의 구현
Cited - time in 
Share 
Share
- Issue Date
- 2010-06
- Citation
- 대한전자공학회 종합 학술 대회 (하계) 2010, pp.467-470
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- We propose the architecture of real-time face detection hardware, and introduce its ASIC implementation. It uses MCT(Modified Census Transform) to obtain face characteristics robustly against illumination changes, and includes Pyramid Image to detect variable size of face. In addition, the AdaBoost algorithm is adopted to learn and generate the characteristics of the face data, and finally to detect the face using this data. This structure was implemented on FPGA(Xilinx Virtex5 LX330), of which maximum speed is 149 fps(frame per second) with QVGA input image. Futhermore, it was integrated a single ASIC(Application-specific integrated circuit) chip named ALTAIR that has the following features ; the size of 1.2㎝ × 1.2㎝, the type of BGA, the speed of 30fps, the detecting accuracy for frontal face of 90%, and the power consumption of 226㎽.
- KSP Keywords
- ASIC implementation, AdaBoost Algorithm, Application Specific Integrated Circuit, Census Transform, Detecting accuracy, Power Consumption, Pyramid image, Real-Time Face Detection, Xilinx virtex, frontal face, illumination change