Implementação, avaliação de desempenho e análise de complexidade computacional de um sistema de cancelamento adaptativo de ruído

Abstract

Nowadays, adaptive noise cancelling (ANC) systems have been used in many different practical applications, such as in noise suppression in hearing aids, audio-conference devices and cellphones. In these applications, since sources of noise can change over time, adaptive filtering techniques are used to synthesize (in real-time) an estimate of noise that corrupts the signal of interest to, then, mitigate it (by subtraction) in the electrical domain. However, only a few efforts have been invested on the practical implementation of adaptive algorithms, on methodologies for performance comparison (with respect to the achieved noise attenuation), as well on computational complexity analyzes. In this context, the present research work deals with implementation, performance evaluation, and computational complexity analysis of an ANC system operating with relevant adaptive algorithms from the literature. Such implementation was carried out in C language using the Cortex FM4 Starter development board from Cypress Semiconductors®, which has acquisition, synthesis, and signal processing resources. In order to assess the performance of the implemented algorithms (vis-a-vis speech signals corrupted by noise), objective methodologies of quality and intelligibility were used. Still, based on the execution time of each function used, a comparison between the computational complexity of the algorithms is presented, thus evidencing computational requirements and important characteristics of the algorithms considered.