基于关注度的可分级音频编码方法研究

项目名称： 基于关注度的可分级音频编码方法研究

项目编号： No.61201247

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 杭波

作者单位： 湖北文理学院

项目金额： 25万元

中文摘要： 公共安全突发事件频发使我国对安防监控需求日益迫切，音频监控不受角度、光照及遮挡限制，是不可或缺的关键支撑技术。监控网络带宽波动频繁，严重影响监控音频质量。可分级音频编码是带宽波动情况下提升音质的最有效手段，但当前基于能量、信掩比等人耳感知特性的可分级音频编码，与安防监控关注音频能量、方位快速变化等特征不吻合，导致分级增强层被抛弃时关注音频信号质量下降。 针对此问题，本课题研究基于心理选择性关注机制的可分级音频编码方法，提升安防监控关注音频质量：1针对传统关注度模型无法表达音源方位快变的问题，引入空间线索变化度量音源方位快变关注度，建立安防监控音频关注度模型；2依据关注失真更易察觉的原理，研究基于关注度失真的带宽可分级编码方法，和基于关注度优先级的信噪比可分级编码方法。本研究对探索基于心理因素的高效音频压缩编码算法具有重要研究价值，对提高安防监控主观音质，促进音频监控技术进步具有重要意义。

中文关键词： 安防监控；音频编码；关注度模型；空间线索；可分级

英文摘要： The frequent occurrence of the public safety emergencies makes the demands for security surveillance increasingly urgent, audio surveillance not limited by the surveillance angle, lighting, and hidden block, is an indispensable security surveillance key supporting technology. Monitor network bandwidth fluctuates frequently, which seriously impacts surveillance audio quality. Scalable audio coding is the most effective means to enhance sound quality under bandwidth fluctuations. But the scalable audio coding based on human perception such as energy, signal-mask-ratio etc. doesn't match the characteristics of attention audio such as energy, location rapidly changing in security surveillance, which result the decline in the quality of the attention audio signal when the scalable enhancement layer is abandoned. To solve this problem, the subject is going to research the scalable audio coding methods which are based on mental selective attention mechanism to enhance the quality of the attention audio in security surveillance: 1.For the traditional audio attention model can't present the fast changing of sound source location, we introduce the spatial cues changing to measure the attention level of the rapidly varying of the sound source location, and establish the audio attention model of the security surveillance; 2

英文关键词： Surveillance；Audio Coding；Attention Model；Spatial Cues；Scalable