Audio Boundary Estimation Model for Consecutive Period Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for estimating consecutive periods of continuous chords in audio signals often result in incorrect estimations due to inaccuracies in chord estimation, affecting the accuracy of subsequent consecutive period analysis.
Innovation Solution
A computer-implemented audio analysis method that acquires feature amounts from audio signals and uses a boundary estimation model, trained on relationships between feature amounts and boundary data, to generate boundary data representing the boundaries between consecutive periods with continuous chords, allowing for high-accuracy estimation of consecutive periods without relying on chord estimation results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If chord estimation techniques are used to estimate consecutive periods, then chord information can be obtained, but incorrect chord estimation causes incorrect consecutive period estimation
Solution Approach 1:
The patent segments the audio analysis task into two independent parts: chord estimation and consecutive period estimation. By using separate models for each task, errors in chord estimation do not propagate to consecutive period estimation, thus resolving the reliability issue.
Solution Approach 2:
The patent introduces feature amounts as an intermediary representation that captures audio characteristics without requiring explicit chord identification. The boundary estimation model operates on these feature amounts directly, avoiding the need for accurate chord estimation while still achieving accurate consecutive period detection.
2Measurement precision
If a boundary estimation model is used to directly estimate consecutive periods from feature amounts, then consecutive period accuracy improves, but the system complexity increases
Solution Approach 1:
The patent uses a neural network model that learns to map audio feature amounts directly to boundary predictions, copying the essential temporal boundary information from the raw audio features without requiring intermediate musical interpretation layers. This simplifies the overall system while maintaining high accuracy.
Data Source
AI summary
An apparatus for analyzing audio includes a feature acquirer, an output apparatus, at least one memory, and at least one processor. The at least one processor is configured to execute a program stored in the at least one memory. The at least one processor is also configured to cause the feature acquirer to acquire a series of feature amounts of an audio signal. The at least one processor is also configured to generate boundary data by inputting the acquired series of feature amounts into a boundary estimation model that has learned relationships between: (i) a series of feature amounts, and (ii) boundary data representative of boundaries, each of the boundaries being between consecutive periods in each of which a chord is continuous.