Score Following via 2D State Space Tempo Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing score following systems are not robust enough to handle quick tempo changes and mistakes during music performance, especially with polyphonic audio inputs and audio interference, leading to inaccurate tracking of musical scores.
Innovation Solution
A two-dimensional state space model is used to analyze audio inputs, where coordinates represent both the location in the score and the tempo, with a Hidden Markov Model and particle filtering process to compute probability distributions and match the performance to the score, allowing for accurate tracking and generation of accompaniment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional score following methods are used, then the system can track musical scores, but the reliability deteriorates when handling quick tempo changes, mistakes, polyphonic inputs, and audio interference
Solution Approach 1:
The patent transitions from traditional one-dimensional score following to a two-dimensional state space that jointly models both location and tempo. This dimensional expansion allows the system to simultaneously track positional progress and tempo variations, resolving the contradiction by providing a more comprehensive state representation that maintains reliability under varying performance conditions
Solution Approach 2:
The patent changes the parameter representation by using continuous tempo as a state variable rather than discrete tempo values. This continuous parameter approach allows the Hidden Markov Model to smoothly track tempo fluctuations and Quick tempo changes, improving reliability without being disrupted by abrupt temporal variations or performance mistakes
2Measurement precision
If complex processing is applied to handle polyphonic inputs and audio interference, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
The patent introduces a Hidden Markov Model as an intermediary computational framework that mediates between the complex polyphonic audio input and the score tracking objective. The HMM with its two-dimensional state space acts as a intermediary layer that processes temporal and spectral features, achieving high measurement precision through probabilistic inference while maintaining manageable system complexity through modular architecture
Solution Approach 2:
The patent segments the score following problem into distinct state variables (location and tempo) and processes them separately within the HMM framework. This segmentation allows the system to handle polyphonic inputs and audio interference by independently tracking positional progress and tempo variations, improving measurement precision without requiring overly complex integrated processing
Data Source
AI summary
A method for audio processing includes receiving in an electronic processor an audio input from a performance of a musical piece having a score. A two-dimensional state space is defined, including coordinates modeling the performance, each coordinate corresponding to a respective location in the score and a tempo of the performance. For each of a plurality of times during the performance, a probability distribution is computed over the two-dimensional state space based on the audio input. Based on the probability distribution, the performance is matched to the score.