|
|
|
|
|
|
NAIST Music Rendering Model The NAIST Music Rendering Model uses a Gaussian statistical machine learning method for generating an expressive music performance and requires that only a few parameters be set manually. The model learns the musical performance so as to predict a musician’s performance, which includes conscious and unconscious variations on the musical score. The musical score information is given as input, and the corresponding actual performance - dynamics (note strength), attack time, and release time - is given as output. The model can directly approximate input-output relationships that might include rules that are hard to describe explicitly. It should be able to imitate a master performer if it is trained using a large enough dataset. YQX The expressive rendering strategy of YQX consists of two parts. Firstly, a reference tempo and a loudness curve are constructed, comprising any hints in the score that concern changes in tempo and loudness (crescendos, ritards etc.). Secondly, YQX comprises a bayesian network that models dependencies of timing, loudness, and articulation of notes on their score contexts. The score contexts are described in terms of features like pitch interval, rhythmic context, and features related to the implication-realization analysis (Narmour, 1990) of the melody. To this end, a simple soprano voice extraction method, and an implication-realization parser (Grachten, 2006) are used. The network has been trained on performances of Mozart's piana sonatas by Batik, and of Chopin's piano works by Magaloff. The final rendering of a piece is obtained by combining the reference tempo and loudness curves and the note-wise predictions of the network. Finally, instanteneous effects such as accent, and fermata are applied where marked in the score. COPER "COPER on the Web" is a Web-based performance rendering system using a corpus of expressive music performances. This system generates expression for a given musical piece by applying expressions in the musical segment extracted from a corpus. The segments are extracted using queries related to the "musical group boundary" or "pitch, duration and transition of the notes" in a melody sequence. The corpus consists of a set of cases (actual examples) of human musical performances, described using descriptors of nominal (quantized) note information and musical deviation terms. The deviation terms consist of whole tempo transition, damper pedal on/off action with depth degree value, and delicate temporal and dynamics tactus control. Kagurame-Phase II Kagurame Phase-II is a performance rendering system that can generate various types of performances for a single piece depending on the parameter settings. This is made possible by using a case-based approach rather than a rule-based one for generating musical expressions. A large amount of human performance data is used as a knowledge base for musical expression generation instead of rules or methods for musical expression. The parameter settings are called the "performance condition." It can include various types of parameters such as style, mood, and performer. (Cancelled) |
ESP The Expression Synthesis Project, ESP, explores the use of the driving metaphor for expressive performance. The ESP system enables a user to interactively create a performance while navigating a road that suggests an interpretation based on musical structure. The "driver" controls a "car" while road bends encourage slow-downs and straight sections encourage speedups. A dashboard shows a roadmap and speedometer (in bpm); buttons control the articulation and (de)coupling of dynamics with acceleration. The system enables users to experiment with performances prior to mastering an instrument or composition and enables composers to affect performance decisions by, for example, crafting a roadmap. iFP M.J. iFP is an interface for playing expressive music. It promotes a pianist's expressiveness with its tapping-style input. MIDI-formatted expressive piano performances were first analyzed and transformed into performance templates in which deviations from the canonical description were separately described for each event. Using one of the templates as a skill complement, a musician can expressively play a musical piece over and under the beat level. The iFP scheduler enables players to mix their own intentions and expressiveness into the performance template. An iFP performer can experience the thrill of virtuosity by varying weight parameters dynamically to control the deviations in tempo, dynamics, and delicate nuances within the beat. That is, iFP provides functions for controlling the intention level of the player and the model performance. It also provides a morphing function that interpolates and extrapolates two different expressive performance templates of a musical piece. The peripheral interface can be a PC keyboard, a musical keyboard, or a conducting interface using capacity transducers. Finale 2008 |
