CISE Research Instrumentation: Multiprocessor Workstation for Advanced Digital Signal Processing Research

CDA-9529520巴雷托，Armando B.Yen，康佛罗里达国际大学cise研究仪器：用于高级数字信号处理研究的多处理器工作站新兴的数字信号处理技术正被应用于解决信号检测和系统建模的挑战性问题：动态神经网络被用作时空分类器，用于检测作为运动准备过程的一部分，可以自愿诱发的脑电模式。对这些模式的可靠检测将为开发一种新形式的计算机接口提供基础，即使是被剥夺了运动执行功能的个人也可以使用这种接口。同样，混沌动力学分析也被应用于详细研究海洋声波的三维传播。利用在北大西洋收集的实际声音传播数据，对诸如Lyapunov指数等指数进行了数值估计，以表征该现象的混沌特性。虽然这些新兴的数字信号处理技术极有希望提高我们从信号中获取有用信息和开发物理现象的准确模型的能力，但它们很复杂，计算密集。为了缩短算法开发和分析周期(即大规模动态神经网络训练、数值Lyapunov指数估计等)，多处理器工作站的可用性是必要的。到实际的水平。

CDA-9529520 Barreto, Armando B. Yen, Kang Florida International University CISE Research Instrumentation: Multiprocessor Workstation for Advanced Digital Signal Processing Research Emerging Digital Signal Processing techniques are being applied to the solution of challenging problems of signal detection and system modeling: Dynamic neural networks are used as spatio-temporal classifiers for the detection of electroencephalographic (EEG) patterns that can be elicited voluntarily, as part of the process of motion preparation. Reliable detection of these patterns will provide the basis for the development of a new form of computer interface that will be accessible even to individuals deprived of motor execution functions. Similarly, chaotic dynamics analysis is being applied for the detailed study of the three-dimensional propagation of ocean acoustic waves. Indices such as the Lyapunov exponent are numerically estimated using real sound propagation data collected in the North Atlantic to characterize the chaotic properties of the phenomenon. While these emerging DSP techniques hold great promise to increase our capability to obtain useful information from signals and develop accurate models of physical phenomena, they are complex and computationally intensive. The availability of multiprocessor workstations is necessary to shorten the algorithm development and analysis cycles (i.e., large-scale dynamic neural network training, numerical Lyapunov exponent estimation, etc.) to practical levels. Additionally, these systems will familiarize our faculty and students with parallel algorithm development techniques