This project covers the analysis of linear prediction coding, and sparse linear prediction algorithms, with the goal of creating a Register Transfer Level architecture. Two algorithm is then suggested to find a set of, linear prediction coefficients, and sparse linear prediction coefficients, these algorithms are analysed finding the flow, and inherent parallelism. With the analysis of these algorithms a finite state machine with data path, is build consisting of a set of the control path, and data path. The control path is build using algorithmic state machine charts, and data path is build consisting a set of hardware blocks.
The sparse linear prediction algorithm analysed in this report is a novel idea recently presented by the supervisor Tobias Lindstrøm Jensen, therefore a larger part of the analysis is used to investigate it. The given algorithm consists of four iterations where a right hand side least squares problem is solved each iteration. Given this it is found that the Levinson algorithm is the most sufficient way of finding a solution, where other methods are also investigated.
The report ends up with two architectures, which can be either implemented in VHDL or further improved employing different optimizations methods.