改进YOLOv7-tiny的手语识别算法研究

Research on sign language recognition algorithm based on improved YOLOv7⁃tiny

在与听障人士进行交流时,常常会面临交流不便的困难,文中提出一种手语识别的改进模型来解决这个困难。该模型基于YOLOv7-tiny网络模型,并对其进行了多项改进,旨在提高模型的精度和速度。首先,通过对CBAM注意力机制的通道域进行改进,解决了因降维而造成的通道信息缺失问题,并将改进后的CBAM加入到YOLOv7-tiny的Neck层中,从而使模型更加精准地定位和识别到关键的目标;其次,将传统的CIoU边界框损失函数替换为SIoU边界框损失函数,以加速边界框回归的同时提高定位准确度;此外,为了减少计算量并加快检测速度,还将颈部层中的普通卷积模块替换为Ghost卷积模块。经过实验测试,改进后网络模型的平均精度均值(mAP)、精准率和召回率分别提升了5.31%、6.53%、2.73%,有效地提高了手语识别网络的检测精确度。

When people need to interact with deaf individuals,they often face barriers in communication.To address this issue,an improved sign language recognition network model based on the YOLOv7⁃tiny has been proposed.The improvements aim at enhancing the accuracy and speed of the model.The channel domain of the CBAM attention mechanism is improved to eliminate the channel information loss caused by dimension reduction.The improved CBAM is added to the Neck layer of YOLOv7⁃tiny to enable the model to locate and recognize the key targets more accurately.The traditional CIoU boundary box loss function is replaced by the SIoU boundary box loss function to improve localization accuracy while accelerating boundary box regression.In addition,to reduce computation and accelerate detection speed,the normal convolution modules in the Neck layer are replaced with Ghost convolution modules.Experimental tests have shown that the mean average precision(mAP),precision rate,and recall rate of the improved network model have increased by 5.31%,6.53%and 2.73%,respectively,effectively improving the detection accuracy of the sign language recognition network.