Robust Automatic Recognition of Speech with Background Music

Simple item page
dc.contributor.authorMálek Jiřícs
dc.contributor.authorŽďánský Jindřichcs
dc.contributor.authorČerva Petrcs
dc.date.accessioned2018-09-25T12:15:05Z
dc.date.available2018-09-25T12:15:05Z
dc.date.issued2017cs
dc.description.abstractThis paper addresses the task of Automatic Speech Recognition (ASR) with music in the background, where the accuracy of recognition may deteriorate significantly. To improve the robustness of ASR in this task, e.g. for broadcast news transcription or subtitles creation, we adopt two approaches: 1) multi-condition training of the acoustic models and 2) denoising autoencoders followed by acoustic model training on the preprocessed data. In the latter case, two types of autoencoders are considered: the fully connected and the convolutional network. Presented experimental results show that all the investigated techniques are able to improve the recognition of speech distorted by music significantly. For example, in the case of artificial mixtures of speech and electronic music (low Signal-to-Noise Ratio (SNR) of 0 dB), we achieved absolute improvement of accuracy by 35.8%. For real-world broadcast news and a high SNR (about 10 dB), we achieved improvement by 2.4%. The important advantage of the studied approaches is that they do not deteriorate the accuracy in scenarios with clean speech (the decrease is about 1%).en
dc.format.extent5cs
dc.identifier.doi10.1109/ICASSP.2017.7953150
dc.identifier.isbn978-1-5090-4117-6cs
dc.identifier.issn1520-6149cs
dc.identifier.urihttps://dspace.tul.cz/handle/15240/31348
dc.identifier.urihttps://ieeexplore.ieee.org/document/7953150
dc.language.isoengcs
dc.publisherInstitute of Electrical and Electronics Engineers Inc.cs
dc.publisher.cityUSAcs
dc.relation.ispartofseries0cs
dc.subjectRobust recognitioncs
dc.subjectbackground musiccs
dc.subjectfeature enhancementcs
dc.subjectdenoising autoencodercs
dc.subjectmulti-condition trainingcs
dc.titleRobust Automatic Recognition of Speech with Background Musicen
dc.titleRobust Automatic Recognition of Speech with Background Musiccs
local.citation.epage5210-5214cs
local.citation.spage5210-5214cs
local.identifier.publikace4811
local.identifier.wok414286205074en
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
ROBUST AUTOMATIC RECOGNITION OF SPEECH WITH BACKGROUND MUSIC.pdf
Size:
140.79 KB
Format:
Adobe Portable Document Format
Description:
článek
Download
Collections