A kind of turbo joint detection scheme based on parallel interference cancellation （PIC） is studied; then, the eigenvalues of iteration matrix is deeply analyzed for studying the ping-pong effects in PIC JD and the ...A kind of turbo joint detection scheme based on parallel interference cancellation （PIC） is studied; then, the eigenvalues of iteration matrix is deeply analyzed for studying the ping-pong effects in PIC JD and the corresponding compensation approach is introduced. Finally, the proposed algorithm is validated through computer simulation in TDD CDMA uplink transmission. The result shows that the ping-pong effects are almost avoided completely in the presence of the compensation scheme, and system performance is greatly improved.展开更多
As the combining form of the orthogonal frequency-division multiplexing (OFDM) technique and the vertical Bell Labs layered space-time (V-BLAST) architecture, the V-BLAST OFDM system can better meet the demand of next...As the combining form of the orthogonal frequency-division multiplexing (OFDM) technique and the vertical Bell Labs layered space-time (V-BLAST) architecture, the V-BLAST OFDM system can better meet the demand of next-generation (NextG) broadband mobile wireless multimedia communications. The symbols detection problem of the V-BLAST OFDM system is investigated under the frequency-selective fading environment. The joint space-frequency demultiplexing operation is proposed in the V-BLAST OFDM system. Successively, one novel half-rate rotational invariance joint space-frequency coding scheme for the V-BLAST OFDM system is proposed. By elegantly exploiting the above rotational invariance property, we derive one direct symbols detection scheme without knowing channels state information (CSI) for the frequency-selective V-BLAST OFDM system. Extensive simulation results demonstrate the validity of the novel half-rate rotational invariance joint space-frequency coding scheme and the performance of the direct symbols detection scheme.展开更多
为实现群养生猪在不同场景下(白天与黑夜,猪只稀疏与稠密)的猪只个体准确检测与实时跟踪,该研究提出一种联合检测与跟踪(Joint Detection and Embedding,JDE)模型。首先利用特征提取模块对输入视频序列提取不同尺度的图像特征,产生3个...为实现群养生猪在不同场景下(白天与黑夜,猪只稀疏与稠密)的猪只个体准确检测与实时跟踪,该研究提出一种联合检测与跟踪(Joint Detection and Embedding,JDE)模型。首先利用特征提取模块对输入视频序列提取不同尺度的图像特征,产生3个预测头,预测头通过多任务协同学习输出3个分支,分别为分类信息、边界框回归信息和外观信息。3种信息在数据关联模块进行处理,其中分类信息和边界框回归信息输出检测框的位置,结合外观信息,通过包含卡尔曼滤波和匈牙利算法的数据关联算法输出视频序列。试验结果表明,本文JDE模型在公开数据集和自建数据集的总体检测平均精度均值(mean Average Precision,m AP)为92.9%,多目标跟踪精度(Multiple Object Tracking Accuracy,MOTA)为83.9%,IDF1得分为79.6%,每秒传输帧数(Frames Per Second,FPS)为73.9帧/s。在公开数据集中,对比目标检测和跟踪模块分离(Separate Detection and Embedding,SDE)模型,本文JDE模型在MOTA提升0.5个百分点的基础上,FPS提升340%,解决了采用SDE模型多目标跟踪实时性不足问题。对比TransTrack模型,本文JDE模型的MOTA和IDF1分别提升10.4个百分点和6.6个百分点,FPS提升324%。实现养殖环境下的群养生猪多目标实时跟踪,可为大规模生猪养殖的精准管理提供技术支持。展开更多
文摘A kind of turbo joint detection scheme based on parallel interference cancellation （PIC） is studied; then, the eigenvalues of iteration matrix is deeply analyzed for studying the ping-pong effects in PIC JD and the corresponding compensation approach is introduced. Finally, the proposed algorithm is validated through computer simulation in TDD CDMA uplink transmission. The result shows that the ping-pong effects are almost avoided completely in the presence of the compensation scheme, and system performance is greatly improved.
文摘As the combining form of the orthogonal frequency-division multiplexing (OFDM) technique and the vertical Bell Labs layered space-time (V-BLAST) architecture, the V-BLAST OFDM system can better meet the demand of next-generation (NextG) broadband mobile wireless multimedia communications. The symbols detection problem of the V-BLAST OFDM system is investigated under the frequency-selective fading environment. The joint space-frequency demultiplexing operation is proposed in the V-BLAST OFDM system. Successively, one novel half-rate rotational invariance joint space-frequency coding scheme for the V-BLAST OFDM system is proposed. By elegantly exploiting the above rotational invariance property, we derive one direct symbols detection scheme without knowing channels state information (CSI) for the frequency-selective V-BLAST OFDM system. Extensive simulation results demonstrate the validity of the novel half-rate rotational invariance joint space-frequency coding scheme and the performance of the direct symbols detection scheme.
文摘为实现群养生猪在不同场景下(白天与黑夜,猪只稀疏与稠密)的猪只个体准确检测与实时跟踪,该研究提出一种联合检测与跟踪(Joint Detection and Embedding,JDE)模型。首先利用特征提取模块对输入视频序列提取不同尺度的图像特征,产生3个预测头,预测头通过多任务协同学习输出3个分支,分别为分类信息、边界框回归信息和外观信息。3种信息在数据关联模块进行处理,其中分类信息和边界框回归信息输出检测框的位置,结合外观信息,通过包含卡尔曼滤波和匈牙利算法的数据关联算法输出视频序列。试验结果表明,本文JDE模型在公开数据集和自建数据集的总体检测平均精度均值(mean Average Precision,m AP)为92.9%,多目标跟踪精度(Multiple Object Tracking Accuracy,MOTA)为83.9%,IDF1得分为79.6%,每秒传输帧数(Frames Per Second,FPS)为73.9帧/s。在公开数据集中,对比目标检测和跟踪模块分离(Separate Detection and Embedding,SDE)模型,本文JDE模型在MOTA提升0.5个百分点的基础上,FPS提升340%,解决了采用SDE模型多目标跟踪实时性不足问题。对比TransTrack模型,本文JDE模型的MOTA和IDF1分别提升10.4个百分点和6.6个百分点,FPS提升324%。实现养殖环境下的群养生猪多目标实时跟踪,可为大规模生猪养殖的精准管理提供技术支持。
