Musical Score Data Processing with Variability Models

The patent segments the performance data generation process into two distinct models: a first provisional model that learns the performer's basic tendency from performance data, and a second provisional model that generates varied performances by incorporating variability data. This segmentation allows the system to maintain accuracy in representing the performer's style while simultaneously enabling generation of diverse performances with different musical expressions.

The patent introduces variability data as a dynamic element that can be adjusted to generate different performances. By making the performance generation process dynamic and adaptable through variability parameters, the system can reflect changes in performance while maintaining the core performer style, thus resolving the contradiction between accuracy and versatility.

The patent divides the complex task of generating diverse performances into manageable segments: the first provisional model handles learning performer tendencies, while the second provisional model handles generation of varied performances. This segmentation makes the overall complex process more manageable while achieving the goal of diverse performance generation.

The patent introduces variability data as an intermediary element that bridges the gap between simple score data and diverse performance outputs. This intermediary allows the system to maintain relative simplicity in the input while achieving diversity in the output, resolving the contradiction between process simplicity and expression diversity.