Performance Evaluation of A Speech Protected Active Noise Control System


CANLI USTA Ö., DOĞAN H.

International Conference on Artificial Intelligence and Data Processing (IDAP), Malatya, Turkey, 21 - 22 September 2019 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/idap.2019.8875951
  • City: Malatya
  • Country: Turkey
  • Keywords: active noise control, convolutive blind source separation, voice activity detection, BLIND, SEPARATION, VOICE
  • Dokuz Eylül University Affiliated: Yes

Abstract

The auditory awareness of a person wearing headphones for protection against noise is very important, especially in a hazardous environment. To enhance the auditory awareness, different systems that contain the various combinations of Voice Activity Detection (VAD), Blind Source Separation (BSS) and Active Noise Cancellation (ANC) blocks have been proposed in the literature. The merit of these systems lies in their ability to enhance or separate speech signals especially when the noise is dominant, and the success of them highly depend on the performance of each block and the signal to noise ratio (SNR). The aim of this study is to determine the abilities and behaviors of a speech-protected active noise control system especially in noise dominant cases and in a reverberant room. For this purpose, a speech protected active noise control system that is the combination of VAD, Convolutive BSS (CBSS) and ANC blocks is presented and the performance of each block evaluated on a dataset that contains noise and speech signals that are mixed with different SNR levels.