Signal processing algorithms for applications in communication systems

数字信号处理(DSP)领域在过去的几十年里经历了爆炸性的增长。数字信号处理技术已经成为我们日常生活中所需要或遇到的产品和服务不可或缺的一部分。申请者在过去五年中的研究努力取得了一些非常具体的成果，这些成果已与国际科学界分享，包括学术界和工业界，并产生了需要进一步研究的新想法和方向。拟议研究的总体目标是开发有效的算法，并为语音、图像和视频信号的可靠处理及其多媒体通信应用的有效软件和硬件实现奠定基本的数学基础。在图像的采集、压缩、存储、传输和再现过程中，加性或乘性噪声引起的失真是很常见的。噪声的存在使得图像数据的自动解释变得非常困难。语音活动检测、噪声污染语音的基音估计以及噪声语音系统的识别和建模对于语音通信系统的可靠设计至关重要。新兴的通信和信息网络处理大量的数字语音、图像和视频信号，并通过过度拥堵的通信信道进行传输。这些活动的基础是信号压缩以及这种信号的前后处理。将努力通过设计和实现性价比高、性能优越的DSP算法来解决这些问题。离散变换对于信号处理任务是必不可少的。为此，将研究更有效地利用信号固有特性的新变换。将采用快速高效的技术来计算这些变换和现有的变换。研究工作还将针对本提案中开发的算法的高效软件和VLSI实现。

The field of digital signal processing (DSP) has experienced an explosive growth during the past couple of decades. The DSP techniques have become integral parts of the products and services that we need or encounter in our daily lives. The research efforts of the applicant during the past five years have led to some very concrete results that have been shared with international scientific community, both from academia and industry, and have given rise to new ideas and directions that need to be further investigated. The overall objective of the proposed research is to develop efficient algorithms and to establish the underlying mathematical foundations for reliable processing of speech, image and video signals, and for their efficient software and hardware implementation for multimedia communication applications. The distortion of images by additive or multiplicative noise is common during their acquisition, compression, storage, transmission and reproduction. Presence of the noise makes the automatic interpretation of image data very difficult. Voice activity detection, pitch estimation of noise-corrupted speech, and identification and modeling of noisy speech systems are vital for reliable design of speech communication systems. Emerging communication and information networks deal with high volumes of digital speech, image and video signals and with their transmission through over-congested communication channels. Fundamental to these activities are the signal compression, and pre- and post-processing of such signals. Efforts will be put to reach solutions to these problems through design and implementation of DSP algorithms that are cost effective and are superior in performance. Discrete transforms are essential to signal processing tasks. To this end, new transforms exploiting more efficiently the intrinsic properties of signals will be investigated. Fast and efficient techniques for computing these and the existing transforms will be undertaken. Research efforts will also be directed towards efficient software and VLSI implementations of the algorithms developed in this proposal.