Action recognition is an important topic in computer vision. Recently, deep learning technologies have been successfully used in lots of applications including video data for sloving recognition problems. However, mos...Action recognition is an important topic in computer vision. Recently, deep learning technologies have been successfully used in lots of applications including video data for sloving recognition problems. However, most existing deep learning based recognition frameworks are not optimized for action in the surveillance videos. In this paper, we propose a novel method to deal with the recognition of different types of actions in outdoor surveillance videos. The proposed method first introduces motion compensation to improve the detection of human target. Then, it uses three different types of deep models with single and sequenced images as inputs for the recognition of different types of actions. Finally, predictions from different models are fused with a linear model. Experimental results show that the proposed method works well on the real surveillance videos.展开更多
针对中文儿童语音情感识别的准确性问题,提出了一种结合深度卷积神经网络(Deep Convolutional Neural Network,DPCNN)与堆叠长短时记忆(Stacked Long Short Term Memory,SLSTM)网络的融合模型,旨在提高中文儿童语音情感识别的准确性。通...针对中文儿童语音情感识别的准确性问题,提出了一种结合深度卷积神经网络(Deep Convolutional Neural Network,DPCNN)与堆叠长短时记忆(Stacked Long Short Term Memory,SLSTM)网络的融合模型,旨在提高中文儿童语音情感识别的准确性。通过DPCNN对语音信号中的长距离依赖关系进行提取,再利用SLSTM捕捉情感相关的序列依赖信息,最终通过softmax分类器实现情感状态的判别。实验结果显示,基于DPCNN-SLSTM的模型在中文儿童语音数据集上的情感识别准确率达到了92%,显著优于CNN、LSTM和CNN-LSTM模型。研究结果对于推动儿童语音情感识别技术的发展具有重要意义。展开更多
文摘Action recognition is an important topic in computer vision. Recently, deep learning technologies have been successfully used in lots of applications including video data for sloving recognition problems. However, most existing deep learning based recognition frameworks are not optimized for action in the surveillance videos. In this paper, we propose a novel method to deal with the recognition of different types of actions in outdoor surveillance videos. The proposed method first introduces motion compensation to improve the detection of human target. Then, it uses three different types of deep models with single and sequenced images as inputs for the recognition of different types of actions. Finally, predictions from different models are fused with a linear model. Experimental results show that the proposed method works well on the real surveillance videos.
文摘针对中文网络安全领域缺乏公开数据集和有效的命名实体识别(Named Entity Recognition,NER)方法,提出一种融合汉字多源信息的网络安全NER方法。通过构建数据集中所有字符的偏旁和字频向量表,增强了中文字向量的特征表达能力,嵌入到改进的词汇融合模型中进行字向量与词向量的融合,输入到条件随机场(Conditional Random Fields,CRF)进行解码。实验结果表明,该方法在保持较快解码速度和占用较低计算机资源的情况下,在网络安全数据集上,其准确率、召回率和F1值分别为0.8649、0.8402和0.8523,均优于现有模型,能够为后续网络安全知识图谱的构建提供支撑。
文摘针对中文儿童语音情感识别的准确性问题,提出了一种结合深度卷积神经网络(Deep Convolutional Neural Network,DPCNN)与堆叠长短时记忆(Stacked Long Short Term Memory,SLSTM)网络的融合模型,旨在提高中文儿童语音情感识别的准确性。通过DPCNN对语音信号中的长距离依赖关系进行提取,再利用SLSTM捕捉情感相关的序列依赖信息,最终通过softmax分类器实现情感状态的判别。实验结果显示,基于DPCNN-SLSTM的模型在中文儿童语音数据集上的情感识别准确率达到了92%,显著优于CNN、LSTM和CNN-LSTM模型。研究结果对于推动儿童语音情感识别技术的发展具有重要意义。