Nombre: HIGOR ARAÚJO FIM CAMPOREZ
Fecha de publicación: 28/05/2020
Supervisor:
Nombre |
Papel |
|---|---|
| HELDER ROBERTO DE OLIVEIRA ROCHA | Advisor * |
| JAIR ADRIANO LIMA SILVA | Co-advisor * |
Junta de examinadores:
| Nombre |
Papel |
|---|---|
| ARNALDO GOMES LEAL JÚNIOR | Internal Examiner * |
| EVANDRO MENARA MILETTO | External Examiner * |
| HELDER ROBERTO DE OLIVEIRA ROCHA | Advisor * |
| JAIR ADRIANO LIMA SILVA | Co advisor * |
Sumario: Musical activities as a tool for fun, entertainment, social interaction, creative expressiveness, and others, have become allies of the human being in their daily lives. Technological advances have provided the creation of tools that support musical activities, for example, the computer. The intersection between technology and music provided the creation of new research areas such as ubimus (ubiquitous music) that uses ubiquitous systems with human agents and material resources (tools to support creativity) for musical purposes. There is also the area named IoMusT (Internet of Musical Things) that provides ubimus applications in IoT (Internet of Things) infrastructure. These areas intersect with robotic musical instruments, which can be controlled by humans, through interaction interfaces, or operate in autonomous mode. This paper aims to develop a platform for networked robotic musical performances through interaction interfaces to control the robots, centralizing efforts on synchronization strategies, in addition, this paper shows the robotic bongo, named BongoBot, development. The platform uses the musical communication protocol Open Sound Control (OSC) to define all message formats. Where the structure of the format has been thought to not limit future generations of robots. As the focus is on synchronization, macro-synchronization algorithms are proposed using SNTP for clocks synchronization and neural networks for training and compensation of the mechanical delays contained in robots. For micro-synchronization, an autonomous perceptual module was implemented, using real-time audio processing to allow the robot to hear the performance and synchronize rhythmically. Tests for macro-synchronization was done with two robots, the neural networks presented an RMSE error of 0.82% and 1.2%. This strategy also provided a reduction of up to 55.33% in the average performance delay and kept all the beats delay of the robots below 20ms (threshold of human sound perception). Micro-synchronization (autonomous perceptual module) was applied to the BongoBot for rhythmic synchronization with a metronome. The numerical evaluation showed that 80% of the tests had an average delay below 20ms and the perceptual evaluation, developed by an online survey, described that 55% of the tests had above 70% of positive votes, that is, votes that the participants agrees with the synchronization.
Keywords: Networked Musical Performance. Robots Synchronization. Ubimus. Internet of Musical Things. Mechanical Delay Compensation by Neural Network.
