Audio Note Detection via Time-Domain Edge Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic transcription technologies for music are unreliable in processing multiple notes simultaneously and fail to accurately capture the full range of music characteristics, such as pitch, duration, velocity, and pedal operations, leading to substandard recreation of music performances and the inability to effectively convert older audio material into modern formats.
Innovation Solution
The system generates frequency domain representations of audio signals over time, detects edges and notes based on these representations, and uses multiple edge detectors to verify the presence of musical notes, allowing for accurate detection and characterization of notes, including pitch, duration, and velocity, enabling precise transcription of music into MIDI format.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional automatic transcription software is used to convert audio to MIDI, then note pitch and duration can be determined, but the system becomes unreliable when processing multiple notes simultaneously and fails to capture full music characteristics
Solution Approach 1:
The audio signal processing is divided into multiple frequency domain representations, each analyzing specific frequency ranges. The system segments the transcription task into detecting different types of edges (onset, release, pitch changes) in time-domain representations derived from these frequency analyses, allowing reliable simultaneous processing of multiple notes while maintaining precision through specialized detection for each edge type
Solution Approach 2:
The patent introduces time-domain representations as an intermediary between frequency domain analysis and note detection. These intermediate representations make music characteristics such as onset, release, and pitch changes explicitly detectable through edge detection, resolving the contradiction by providing a reliable bridge that preserves measurement precision across multiple notes
2Measurement precision
If manual transcription techniques are used by skilled musicians, then accurate music characteristics can be captured, but the process requires significant time and cannot efficiently convert large amounts of audio material
Solution Approach 1:
The patent replaces the mechanical process of manual listening and transcription with an automated computer-based system that uses frequency domain analysis and edge detection algorithms. This substitution maintains measurement precision by detecting music characteristics through mathematical analysis of time-domain representations, while eliminating the time loss associated with manual transcription of large audio archives
Solution Approach 2:
The system performs automatic transcription without requiring skilled musicians to manually listen and transcribe. The automated detection of edges in time-domain representations enables the system to self-identify note characteristics including pitch, duration, onset, and release, efficiently converting large amounts of audio material while preserving accuracy
3Loss of information
If frequency domain analysis is used to detect notes, then pitch information can be obtained, but the system lacks the ability to accurately detect onset, release, and other temporal characteristics
Solution Approach 1:
The patent transforms frequency domain representations into time-domain representations, adding a temporal dimension to the analysis. This dimensionality change enables simultaneous detection of pitch information (from frequency analysis) and temporal characteristics such as onset, release, and duration (from edge detection in time-domain representations), eliminating information loss while maintaining measurement precision
Data Source
AI summary
Methods, system and/or computer program products for detection of a note include receiving an audio signal and generating a plurality of frequency domain representations of the audio signal over time. A time domain representation is generated from the plurality of frequency domain representations. A plurality of edges are detected in the time domain representation and the note is detected by selecting one of the plurality of edges as corresponding to the note based on characteristics of the time domain representation.