Reconstructing Speech Input from Convolutional Neural Network Activity

Simple item page
dc.contributor.advisorGerven, M.A.J. van
dc.contributor.advisorGüçlü, U.
dc.contributor.authorChurchman, T.J.
dc.date.issued2015-07-13
dc.description.abstractConvolutional Neural Networks (CNNs) applied to the auditory domain have achieved great results. However, little research has been performed to uncover the underlying mechanisms that allow auditory CNNs to reach these achievements. This exploratory research attempts to help uncover these mechanisms by using a CNN's activation patterns generated by speech inputs to reconstruct those inputs. It is found that training a decoder to decode activity patterns to a preliminary reconstruction and consequently fine-tuning that reconstruction through further back propagation generates the best results. The reconstructions show that the network preserves a good representation of the input up to and including the fully connected units. Reinforcingly, this representation appears to be suited to speech, instead of, for example, audio in general. Furthermore, it becomes apparent that the network is insensitive to input intensity as well as to the input's activation scale on the time-domain. Further research is required to discover more properties of CNNs applied to the auditory domain.en_US
dc.identifier.urihttp://theses.ubn.ru.nl/handle/123456789/259
dc.language.isoenen_US
dc.thesis.facultyFaculteit der Sociale Wetenschappenen_US
dc.thesis.specialisationBachelor Artificial Intelligenceen_US
dc.thesis.studyprogrammeArtificial Intelligenceen_US
dc.thesis.typeBacheloren_US
dc.titleReconstructing Speech Input from Convolutional Neural Network Activityen_US
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Churchman, T.,_BA_Thesis_2015.pdf
Size:
719.73 KB
Format:
Adobe Portable Document Format
Description:
Scriptietekst
Download
Collections
Faculteit der Sociale Wetenschappen