摘要
基于实现小样本数据集下手势识别的目的,采用了深度卷积神经网络GoogLeNet模型以及PNN神经网络进行分类,同时结合了迁移学习的方法将深度学习模型进行迁移而构建所用模型。用公共数据集Keck Gesture进行实验,通过对数据集图像进行简单的图像预处理,使得图像特征更为明显,将预处理后的图像作为网络输入进行手势识别实验。经实验验证,该方法在该数据上平均准确率达到了99%以上,而且识别速度较快,达到了10帧/s,基本能满足实时性要求。
In order to achieve the purpose of gesture recognition under a small sample data set,a deep convolutional neural network GoogLeNet model and a PNN neural network are used for classification.At the same time,a deep learning model is migrated with a combination of transfer learning methods to build the model used in this paper.In this paper,the public data set Keck Gesture is used for experiments.Simple image preprocessing is performed on the dataset images to make the image features more obvious.The preprocessed pictures are used as network input for gesture recognition experiments.The experimental verification shows that the average accuracy rate of the method in this paper reaches more than 99%,and the recognition speed reaches 10 frames/s,which can basically meet the real-time requirements.
作者
程冉
史健芳
CHENG Ran;SHI Jianfang(School of Information and Computer Science,Taiyuan University of Technology,Jinzhong 030600,China)
出处
《电子设计工程》
2021年第2期179-184,共6页
Electronic Design Engineering
关键词
卷积神经网络
手势识别
概率神经网络
迁移学习
Convolutional Neural Network
gesture recognition
Probabilistic Neural Network
transfer learning
