ID	原文	译文
4313	为解决轨迹差分隐私保护中存在的隐私预算与服务质量等问题，提出了一种融合预测扰动的轨迹差分隐私保护机制。	To address the issues of privacy budget and quality of service in trajectory differential privacy protection, atrajectory differential privacy mechanism integrating prediction disturbance was proposed.
4314	首先，利用马尔可夫链和指数扰动方法预测满足差分隐私和时空安全的扰动位置，并引入服务相似地图检测该位置的可用性；	Firstly, Markov chain and ex-ponential perturbation method were used to predict the location which satisfies the differential privacy and temporal andspatial security, and service similarity map was introduced to detect the availability of the location.
4315	如果预测成功，则直接采用预测位置替代差分扰动的位置，以降低连续查询的隐私开销并提高服务质量。	If the prediction wassuccessful, the prediction location was directly used to replace the location of differential disturbance, to reduce the pri-vacy cost of continuous query and improve the quality of service.
4316	在此基础上，设计基于 w 滑动窗口的轨迹隐私预算分配机制，确保轨迹中任意连续的 w 次查询满足 ε−差分隐私，解决连续查询的轨迹隐私问题。	Based on this, the trajectory privacy budget allocationmechanism based on w sliding window was designed to ensure that any continuous w queries in the trajectory meet theε-differential privacy and solve the trajectory privacy problem of continuous queries.
4317	此外，基于敏感度地图设计一种隐私定制策略，通过自定义语义位置的隐私敏感度，实现隐私预算的量身定制，从而进一步提高其利用率。	In addition, a privacy customizationstrategy was designed based on the sensitivity map. By customizing the privacy sensitivity of semantic location, the pri-vacy budget could be customized to improve its utilization.
4318	最后，利用真实数据集对所提方案进行实验分析，结果显示所提方案提供了更好的隐私保护水平和服务质量。	Finally, the validity of the scheme was verified by real dataset experiment. The results illustrate that it offers the better privacy and quality of service.
4319	针对语音通话中语音段的起始检测性能不佳，检测语音连续性结构受到破坏的问题，提出了一种基于特征流融合的带噪语音检测算法。	Aiming at the problem that the initial detection performance of voice segment was poor, and the voice continu-ity structure was damaged in voice communication, a noisy voice detection algorithm based on feature stream fusion wasproposed.
4320	首先，根据语音特性分别提取时域特征流、谱图特征流和统计特征流；	Firstly, the time domain feature stream, the spectral pattern feature stream and the statistical feature streamwere extracted according to the voice characteristics.
4321	其次，利用不同的语音特征流分别对带噪音频中的语音段进行概率估测；	Secondly, the voice segment in the noisy audio was estimated bydifferent voice feature streams.
4322	最后，将各个特征流估测得到的语音估测概率进行加权融合，并利用隐马尔可夫模型对语音估测概率进行短时状态处理。	Finally, the voice prediction probability obtained by each feature stream was weightedand fused, and the voice estimation probability was processed in short time by the hidden Markov model.