结合Bi-2DPCA与CNN的美式手语识别  

American Sign Language Recognition Combining with Bi-2DPCA and CNN

作　　者：杨明羽 叶春明[1] YANG Mingyu;YE Chunming(School of Business,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学管理学院,上海200093

出　　处：《计算机工程》2021年第12期278-284,共7页Computer Engineering

基　　金：国家自然科学基金(7184003);上海市科委“科技创新行动计划”软科学重点项目(20692104300);上海理工大学科技发展基金(2018KJFZ043)。

摘　　要：针对现有美式手语(ASL)识别算法准确率低和模型训练时间长的问题,提出一种结合双向二维主成分分析(Bi-2DPCA)与卷积神经网络(CNN)并基于贝叶斯优化的识别算法。利用Bi-2DPCA算法对原始图像做数据降维处理,提取行、列方向的特征图,使用卷积神经网络对特征图进行训练分类,同时采用贝叶斯优化算法对模型超参数进行自动调参。在24分类ASL数据集上的实验结果表明,该算法的识别准确率达到99.15%,训练时间相比传统CNN算法减少90.3%。The existing algorithms for American Sign Language(ASL)recognition are limited in the recognition accuracy,and require much time for model training.To address the problem,a Bayesian Optimization(BO)-based algorithm that combines Bidirectional Two-Dimensional Principal Component Analysis(Bi-2 DPCA)and Convolutional Neural Network(CNN)is used to optimize model parameters.The Bi-2 DPCA algorithm is used to reduce the dimensionality of the original image data,and extract the feature maps in the row and column directions.Then the convolutional neural network is used to train and classify the feature maps.Finally,the Bayesian optimization algorithm is used to adjust the model hyperparameters automatically.The experimental results On 24 classified ASL data sets show that the algorithm achieves a recognition accuracy of 99.15%,and reduces the running time by 90.3%compared with the traditional CNN algorithms.

关 键 词：美式手语识别 双向二维主成分分析 卷积神经网络 贝叶斯优化 自动调参 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]