Design of an FPGA based ambisonic codec

Abstract

Ambisonic is a method for capturing three dimensional sound fields through a small number of microphones. Theoretically, the recorded signals could be reconstructed with arbitrary loudspeaker topologies which are unrelated to the original placement of the microphones. Despite the success of the method, there is a general lack of economical hardware solutions to replay the Ambisonic encoded data. In addition, existing products are designed for reconstructing the sound field with rather restricted loudspeaker configurations, defeating one of the major strength of the technology. There are five major objectives in this project. First, to design and implement a digital Ambisonic decoder on a Spartan III Xilinx FPGA platform. The prototype is equipped with I2S interfaces which can be readily connected to digital source and a four channel Class D Amplifiers. Second, study is conducted on some recent research works which utilize modified Tabu search (proposed by Bruce Wiggins) to generalize the decoding process for irregular loudspeaker configurations. Based on the findings of these works, as well as some of the source codes provided, a simulation program is implemented. Evaluation reveals that the method is only effective for symmetrical loudspeaker topology. In the third objective of this project a novel attempt is made to overcome the above problem by replacing the modified Tabu search with a Heuristic Genetic Algorithm. The result is encouraging, showing that the proposed method is superior to the parent scheme in sound field reconstruction. Fourth, the results obtained by the parent and the proposed method are integrated into the Ambisonic decoder for subjective evaluation. Due to the limited gate-count of the FPGA chip optimization is performed to reduce the number of arithmetic operations. Finally, a rotation mechanism is added to the Ambisonic decoder to replay the sound field to audience at different orientation. This further relaxes the need of a fixed angular relation between the listener and the loudspeakers. To a certain extent all the objectives have been successfully accomplished in this project but there are also plenty of rooms for further investigation and improvement. Some of the major issues have been identified and possible solutions are suggested for future development.