Smitha K G, Vinod A P
School of Computer Engineering, Nanyang Technological University, Singapore, Singapore.
Med Biol Eng Comput. 2015 Nov;53(11):1221-9. doi: 10.1007/s11517-015-1346-z. Epub 2015 Aug 4.
Children with autism spectrum disorder have difficulty in understanding the emotional and mental states from the facial expressions of the people they interact. The inability to understand other people's emotions will hinder their interpersonal communication. Though many facial emotion recognition algorithms have been proposed in the literature, they are mainly intended for processing by a personal computer, which limits their usability in on-the-move applications where portability is desired. The portability of the system will ensure ease of use and real-time emotion recognition and that will aid for immediate feedback while communicating with caretakers. Principal component analysis (PCA) has been identified as the least complex feature extraction algorithm to be implemented in hardware. In this paper, we present a detailed study of the implementation of serial and parallel implementation of PCA in order to identify the most feasible method for realization of a portable emotion detector for autistic children. The proposed emotion recognizer architectures are implemented on Virtex 7 XC7VX330T FFG1761-3 FPGA. We achieved 82.3% detection accuracy for a word length of 8 bits.
患有自闭症谱系障碍的儿童在从与他们互动的人的面部表情中理解情感和心理状态方面存在困难。无法理解他人的情感会阻碍他们的人际交流。尽管文献中已经提出了许多面部情感识别算法，但它们主要是为个人计算机处理而设计的，这限制了它们在需要便携性的移动应用中的可用性。系统的便携性将确保易于使用和实时情感识别，这将有助于在与照顾者交流时立即得到反馈。主成分分析（PCA）已被确定为要在硬件中实现的最不复杂的特征提取算法。在本文中，我们对PCA的串行和并行实现进行了详细研究，以确定实现用于自闭症儿童的便携式情感检测器的最可行方法。所提出的情感识别器架构在Virtex 7 XC7VX330T FFG1761-3 FPGA上实现。对于8位字长，我们实现了82.3%的检测准确率。
