CAJ | 학술논문

为提高人脸表情识别算法的识别率和鲁棒性，本文提出一种融合单演二值编码的人脸表情识别算法。该算法运用单演信号分析提取多尺度单演振幅、相位和方向三个正交互补的分量，使用单演二值编码对该三种分量的每个尺度进行编码及划分为多个矩形块子区域，并采用分块 Fisher线性判别对其降维并提高识别率。实验结果表明：所提算法比传统人脸表情识别算法具有更高的识别率。此外，遮挡对比实验证明了所提算法比传统算法有更好的鲁棒性。
위제고인검표정식별산법적식별솔화로봉성，본문제출일충융합단연이치편마적인검표정식별산법。해산법운용단연신호분석제취다척도단연진폭、상위화방향삼개정교호보적분량，사용단연이치편마대해삼충분량적매개척도진행편마급화분위다개구형괴자구역，병채용분괴 Fisher선성판별대기강유병제고식별솔。실험결과표명：소제산법비전통인검표정식별산법구유경고적식별솔。차외，차당대비실험증명료소제산법비전통산법유경호적로봉성。
To improve the recognition rate and robustness of facial expression recognition, a novel facial expression recognition algorithm based on fusion of monogenic binary coding is proposed. This algorithm extracts three orthogonal complementary component products of multi-scale monogenic amplitude, phase and direction respectively by means of monogenic signal analysis, encodes the dimension of each component product using monogenic binary coding and divide each into multiple rectangular regions, and then reduce their dimension and improve their recognition rate by means of Blocked Fisher liner discrimination. Experiment results show that the proposed algorithm has a better recognition rate than the traditional algorithm. Also, occlusion experiments show that it is of better robustness.