University of Piraeus (UNIPI)
Μεθοδολογίες αναγνώρισης προτύπων για την ανάκτηση πολυμεσικών δεδομένων με βάση το περιεχόμενο και μέσω διαβαθμιστικού λογισμικού
Abstract
dc:descriptionIn this thesis it was analyzed and developed a prototype system with middleware software that offers a mobile service that allows mobile users to make a content based search and retrieval of music files belonging in the same musical genre. For this purpose it was used a combination of technologies of mobile technologies techniques of content based musical genre classification and methodologies of individualization and evaluation. The system architecture is consisted of two levels and uses middleware software for the developing of mobile service. The front end of system was designed and developed taking into consideration the characteristics of environment of mobile services in order to eliminate restrictions that existed in similar systems. In our system the mobile service is offered through the push framework. As a result the service is available in all mobile phones, it does not need installation of special software or it needs some special software or mobile phone operation systems. Moreover, the results of service were not content delivery that might be undesirable from the user but a list of links with music files that belong in the same musical genre giving the ability to user of selecting if he wants to receive the corresponding content in his mobile phone. The back end of the systems consists of all these modules that are responsible for the musical genre classification. In order to achieve better musical genre classification results was proposed a new classification method that was based on the musical source separation technique. In this method the original music signal of music file in a preprocessing level was decomposed to its sources. For the separation of the original signal to the three sources were used two relevant algorithms namely Convolutive Sparse Coding and a Wavelet Packets based algorithm. A feature extraction procedure is applied to each separated signal and the extracted features are fed into an ensemble combination of Support Vector Machine based classifiers for genre classification. Experimental results gave a better classification accuracy almost 2-3% in applying a source separation algorithm before feature extraction. Moreover, it was proposed the use of a hybrid method of individualization that had as result the service to provide results that not only were content similar and belonged in the same musical genre but also results that had taken into account also the preferences of the user or other users. The hybrid method of individualization was constituted by the combination of content based music retrieval and a method of collaborating filtering that is named personality diagnosis. For the evaluation of system by the side of software was proposed a three stage evaluation framework. In this framework is combined subjective evaluation criteria relating to users responses and objective evaluation criteria relating to software parameters.
Degree
thesis:*- Grantor dc:publisher
- University of Piraeus (UNIPI)
- Year dc:date
- 2010
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Lampropoulou, Paraskevi
- Λαμπροπούλου, Παρασκευή
Subjects
dc:subject × 20- Ανάκτηση με βάση το περιεχόμενο
- Πολυμεσικά δεδομένα
- Διαβαθμιστικό λογισμικό
- Μηχανές διανυσματικής υποστήριξης
- Διαχωρισμός πηγών
- Υπηρεσίες κινητής τηλεφωνίας
- Εξατομίκευση
- Αξιολόγηση συστημάτων με διαβαθμιστικό λογισμικό
- Content based retrieval
- Multimedia data
- Middleware software
- Support vector machines
- Source separation
- Mobile services
- Personalization
- Evaluation of middleware systems
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/24701