We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.AS

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Sound

Title: Voice Activity Detection in presence of background noise using EEG

Abstract: In this paper we demonstrate that performance of voice activity detection (VAD) system operating in presence of background noise can be improved by concatenating acoustic input features with electroencephalography (EEG) features. We also demonstrate that VAD using only EEG features shows better performance than VAD using only acoustic features in presence of background noise. We implemented a recurrent neural network (RNN) based VAD system and we demonstrate our results for two different data sets recorded in presence of different noise conditions in this paper. We finally demonstrate the ability to predict whether a person wish to continue speaking a sentence or not from EEG features.
Comments: On preparation for submission to EUSIPCO 2020. arXiv admin note: text overlap with arXiv:1906.08871, arXiv:1909.09132
Subjects: Sound (cs.SD); Audio and Speech Processing (eess.AS); Signal Processing (eess.SP)
Cite as: arXiv:1911.04261 [cs.SD]
  (or arXiv:1911.04261v5 [cs.SD] for this version)

Submission history

From: Gautam Krishna [view email]
[v1] Fri, 8 Nov 2019 07:16:27 GMT (365kb,D)
[v2] Wed, 13 Nov 2019 03:52:00 GMT (365kb,D)
[v3] Tue, 10 Dec 2019 22:55:36 GMT (365kb,D)
[v4] Mon, 16 Dec 2019 04:50:14 GMT (423kb,D)
[v5] Sat, 14 Mar 2020 04:51:06 GMT (423kb,D)

Link back to: arXiv, form interface, contact.