Implementation of non-equal-partition multi-channel convolver

Convolution has become a largely exploited signal operation thanks to his several applications in digital signal processing. In the realm of audio elaboration, convolution has the particular meaning of imposing a spectral and/or temporal structure onto a sound. These structures are completely given by the signal with which the signal is being convolved, called Impulse Response (IR). These signals contain a sort of acoustical footprint that can be completely transferred to another sound, earning the same acoustic characteristics as a consequence. With a multichannel approach, convolution assumes even a further meaning and a wider application field. Indeed, it's exploited to deal with modern spatial sound techniques such as Ambisonics which necessitate matrix elaborations of the involved signals. Ambisonics recordings, for example, are made by special coincident multi-capsule microphone arrays, whose signals can be converted to standard Ambisonics format by a multi-channel convolver. A similar concept can apply to the mixing stage of audio production, where direction-based audio objects must be converted to the Ambisonics format to be reproduced in the relative speaker setups. The aim of the work is to analyse an existing algorithm of a multichannel convolver software evaluating his efficiency. Moreover, the managing of the matrix of filters has showed weaknesses when assembling new matrices. Solution proposes a handy way to deal with matrices and to improve the efficiency of the algorithm.