Concurrent algorithms and pipelined processors via decomposition and data-flow principle for recursive filters

Thesis/Dissertation · Thu Jan 01 04:00:00 EST 1987

OSTI ID:6355123

Kiaei, S

The majority of signal-processing algorithms involve complex numerical matrix algebra, and this requires intensive real-time computing speed. To achieve the speed-up for real-time computations, a parallel-processing environment should be designed. A proper combination of the algorithmic and architectural parallelism is necessary for this processing environment. This thesis deals with the design of a concurrent algorithm and a pipelined architecture for recursive filters. The data-level decomposition method is used to obtain a parallel algorithm for the Kalman filter. This method is based on the multiple-data-processing strategy in which the data are segmented, and the segments are processed in parallel. A new approach for the data-level decomposition of the Kalman filter based on the oblique projection is presented. A new pipelined architecture for the VLSI implementation of recursive filters is developed to achieve higher speed-up. The new architecture, termed independent-data-flow-wavefront-array processors (IDFWAP), is a data-flow machine with a feed-back structure suitable for the recursive filters.

• Research Organization: Washington State Univ., Pullman (USA)
• OSTI ID: 6355123
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990210* -- Supercomputers-- (1987-1989)
ALGORITHMS
COMPUTER ARCHITECTURE
DATA PROCESSING
DIGITAL FILTERS
ELECTRONIC CIRCUITS
INTEGRATED CIRCUITS
MATHEMATICAL LOGIC
MICROELECTRONIC CIRCUITS
PARALLEL PROCESSING
PROCESSING
PROGRAMMING
REAL TIME SYSTEMS
SIGNAL CONDITIONING
VECTOR PROCESSING