Unit:
Κατεύθυνση Ηλεκτρονικός Αυτοματισμός (Η/Α, με πρόσθετη εξειδίκευση στην Πληροφορική και στα πληροφοριακά συστήματα)Library of the School of Science
Supervisors info:
Διονύσης Ρεΐσης, Αναπληρωτής Καθηγητής, Τμήμα Φυσικής, Σχολή Φυσικών Επιστημών
Έκτορας Νισταζάκης, Αναπληρωτής Καθηγητής, Τμήμα Φυσικής, Σχολή Φυσικών Επιστημών
Δημήτριος Φραντζεσκάκης, Καθηγητής, Τμήμα Φυσικής, Σχολή Φυσικών Επιστημών
Original Title:
Encoding of audio information for categorizing audio samples.
Translated title:
Encoding of audio information for categorizing audio samples.
Summary:
With the use of a powerful tool, implemented by Numenta, this thesis tries to implement a system, simulating the mammalian auditory system, that will obtain audio samples and try to extract a comparable representation of it. These representations could then be used in various ways and for many applications. One of them could be the embedment of these representations into an Euclidean space leading to their, corresponding audio samples, catecorization.
To do so, an algorithm simulating the mammalian cochlea and its components is designed. This thesis also makes use of a time to frequency transform specially designed for this purpose.
For the context of this thesis, the system designed is partially implemented and an experiment, acting as a proof of concept, is performed. The results show that the algorithm has an effective performance and leaves many promises to future endeavours.
Main subject category:
Science
Other subject categories:
Technology - Computer science
Keywords:
Neural network, Sound, Cochlea, Encoding, Categorization