容错数字信号处理系统的设计方法及实现

高物可靠性的信号处理对于并行处理器的高速实时应用越来越重要。已知在状态空间模型下线性数字信号处理算法可以表示秋为阵与矢量乘积形式，在该模型下，我们假设同一种运算在主计算过程中具有相同的错误概率，即计算过程是珠，并利用矩阵划分的方法提出一种具有纠错能力的容错设计方案。

DESIGN OF FAULT-TOLERANT DIGITAL SIGNAL PROCESSING SYSTEMS: APPROACH AND IMPLEMENTATION

High reliable signal processing has become more and more important forhigh speed real-time applications on parallel architectures. It is known that lineardigital signal processing algorithm can be represented in the form of state spacemodel as matrix-vector multiplication. In the case of the model, by assuming thatone operation has the same error probability in computational process, i. e., thecomputational process is stationary. This paper proposes a fault-tolerant designscheme with high error correction capability by matrix partitioning. If the matrix isdivided into k parts, the scheme can correct k errors, detect 2k errors, in extremecases. Comparing with previous schemes, the scheme can be easily implemented,and has a lower cost in terms of running time and hardware overhead.