dc.contributor.author |
Qazi, Khurram Ashfaq |
|
dc.date.accessioned |
2019-07-04T06:51:48Z |
|
dc.date.accessioned |
2020-04-11T15:36:02Z |
|
dc.date.available |
2020-04-11T15:36:02Z |
|
dc.date.issued |
2018 |
|
dc.identifier.govdoc |
17378 |
|
dc.identifier.uri |
http://142.54.178.187:9060/xmlui/handle/123456789/5080 |
|
dc.description.abstract |
Since decade, a lot of music fingerprinting and speech segregation algorithms have
exhaled. Music speech segregation includes music identification and followed by
speech segregation. This becomes challenging in the presences of the noisy
environment and noisy sample case. A rapid development has taken place in the field
of multimedia content analysis. Music information retrieval applications increased the
emphases on the development of music fingerprinting algorithms. Noise affects the
efficiency and accuracy of the audio information retrieval algorithms.
This research thesis presents a deep analysis of music fingerprinting and speech
segregation algorithms. A novel algorithm is presented for music fingerprinting which
is used for efficient speech segregation in which music fingerprinting is performed over
a noisy audio sample.
This research work proposes a system that performs music fingerprinting in-depth
evolving the speech segregation processes in presence of background noise. Noise is
removed from the audio signal using layered separation model of the recurrent neural
network. Music fingerprinting is performed on the basis of pitch based acoustic features
classified using distributed dictionary based features learning model. The classified
music is processed for speech segregation after noise removal using layered separation
model. Speech is segregated using vocal based acoustic features. Features are classified
using improved dictionary based fisher algorithm. Structured based classes are used for
the classification process.
The systematic evaluation of the proposed system for music fingerprinting and speech
segregation produces competitive results for three datasets (i.e. TIMIT, MIR-1K, and
MusicBrainz), and the results indicate the strength of the proposed system. The
proposed system produces significantly better results when the qualitative and
quantitative analysis is carried out over the standard datasets showing the better
efficiency of our proposed system from the past systems. |
en_US |
dc.description.sponsorship |
Higher Education Commission, Pakistan |
en_US |
dc.language.iso |
en_US |
en_US |
dc.publisher |
University of Engineering & Technology, Taxila. |
en_US |
dc.subject |
Software Engineering |
en_US |
dc.title |
Efficient Music Fingerprinting For Music Speech Segregation |
en_US |
dc.type |
Thesis |
en_US |