Heterogeneous parallel architecture for improving signal processing performance in Doppler blood flow instrumentation

The work presents a heterogeneous parallel approach, for implementing a model based spectral estimation method, based on the modified covariance parametric algorithm. A Heterogeneous Processing Node (HPN) has been developed, which combines the communications ability of the transputer with a high performance signal processing of a DSP, in order to improve the time performance in parametric spectral estimation. The HPN includes libraries for downloading and executing precompiled ready to run DSP functions, using standard transputer programming tools. The heterogeneous approach has been used in a case study concerning parametric spectral estimation of Doppler ultrasound signals. Performance analysis and results are presented with respect to a homogeneous approach, revealing the effectiveness of the heterogeneous approach.