ID	原文	译文
40606	针对当前神经网络加速器难以高效实现目标跟踪边框后处理的问题，提出一种高效的目标跟踪专用加速器.	Since the current nerual network accelerator couldn't efficiently accelerate the post-processing of object tracking, a dedicated object trackeris proposed.
40607	引入神经网络架构，用于提取输入视图特征	A neural network architecture is introduced to extract the features of the input feature map.
40608	并生成边框置信度与偏移量集合.	At the meanwhile, it generates thebounding box confidence and position offset sets.
40609	随后针对目标跟踪的边框处理设计了专用于边框的回归、惩罚以及提取操作的加速模块，	Adedicated acceleration module is designed for the anchor regression, penalty calculation and extraction.
40610	通过同步神经网络加速器与专用加速模块间的数据，以流水结构并行执行特征提取与边框操作，	By synchronizing the data between the neural network accelerator and the dedicated module, a new pipelined structure is proposed to execute the feature extraction and anchor regression in parallel.
40611	实现基于深度学习目标跟踪的端到端处理.	Therefore, the end-to-end processing of the object tracking is efficiently achieved.
40612	设计了一种低成本的64倍降采样数字抽取滤波器，对∑△ADC的输出码流进行滤波和抽取.	A low-cost 64 times down sampling digital decimation filter is designed to filter and extract the output stream of ∑△ ADC.
40613	为节省面积和保证稳定性，首先选用2抽取级联的滤波器实现方式;	In order to save the area and ensure stability, a two-decimation cascaded filter type is selected firstly.
40614	其次对单级滤波器进行结构优化，采用更省面积的折叠转置结构;	Secondly, the structure of the single-stage filter is optimized, and the folded transposition structure with less area is adopted;
40615	在此基础上对系数相乘与加法部分进行了系数优化和公共项提取;	on this basis, the coefficient optimization and common term extraction of the coefficient multiplication and addition part are carried out.