Πανεπιστήμιο Μακεδονίας
Αλληλεπίδραση ανθρώπου – υπολογιστή για την μετάδοση γνώσης με αξιοποίηση μηχανικής μάθησης και αισθητηριοκινητικών τεχνικών σε πολυτροπικό σήμα».: πεδίο εφαρμογής: Βυζαντινή μουσική
Abstract
dc:descriptionIn this dissertation, an innovative approach for the training of a new chanter (apprentice) through a continuous improvement process of interpretation in relation to the same interpretation of a teacher is presented. The proposed approach utilizes the multimodal – multimedia signal handling (sound and video) and the combined use of innovative appliances for human-computer interaction. Moreover, it includes the formation of a hymn-recognition system based on a collection which has developed from recordings of teachers and apprentices’ interpretations.For the purposes of the proposed approach, an innovative training methodology that utilizes machine learning techniques has been developed. More specifically, the methods used for system’s training are based on modeling of multimodal signal through the use of Hidden-Markov-Models (HMM) and Dynamic Time Wrapping (DTW) techniques. In this way, after the recognition of the apprentice’s interpretation in relation to the corresponding interpretation of the teacher, a comparison between the two hymns is performed based on the musical similarity of their tonality. The signals of the two hymns are temporarily fragmented based on their musical beat characteristic. In order to find the musical beat, suitable for human-computer interaction techniques of computer vision are implemented. Therefore, the perception of the movement of the teacher’s or the apprentice’s right hand leads to the extraction of the musical beat of their interpretation. The results of the two hymns comparison are exported as feedback from the system. In the context of this interaction, the new chanter is being informed about the correctness of his/hers interpretation in relation with the same interpretation of the teacher.The evaluation of the system based on the recognition of the hymns through the use of multimodal signal has been accomplished through a comparative study of related research works and has shown higher percentages of successful recognition of hymns. As far as it concerns the process of comparing two recognized hymns by using computer vision technologies, the positive results coming from the retroaction of the system, regarding the musical similarity of their tonality, confirm the efficient use of the training methodology in order the apprentice’s interpretation to be improved.The proposed training methodology belongs to the domain of human-computer interaction and is expected to contribute significantly in the preservation and spread of our cultural legacy and more specifically of the Byzantine Music (BM).
Degree
thesis:*- Grantor dc:publisher
- Πανεπιστήμιο Μακεδονίας
- Year dc:date
- 2017
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kokkinidis, Konstantinos-Hercules
- Κοκκινίδης, Κωνσταντίνος-Ηρακλής
Subjects
dc:subject × 18- Όραση Η/Υ
- Επεξεργασία πολυτροπικού - πολυμεσικού (ήχος και βίντεο) σήματος
- Κρυφά μαρκοβιανά μοντέλα
- Δυναμική περιτύλιξη χρόνου
- Διασταυρούμενη επικύρωση
- Computer vision
- Multi - modal multimedia (sound and video) signal processing
- Hidden - markov - models (HMM)
- Dynamic time wrapping (DTW)
- Cross validation leave-one or jackknife
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Κοινωνικές Επιστήμες
- Εκπαίδευση
- Natural Sciences
- Computer and Information Sciences
- Social Sciences
- Education
Rights
dc:rights- Statement dc:rights
-
- BY_NC
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/44478