On-Device Music Recognition Using Low-Power Audio Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing devices face challenges in efficiently identifying songs playing in the environment without requiring network connectivity or significant battery consumption, as they often rely on remote server analysis which is slow and computationally expensive.
Innovation Solution
Implementing a system where a low-power processor continuously monitors ambient audio for music presence using machine learning, and a high-power processor identifies songs by comparing audio characteristics with an on-device dataset, allowing for real-time song recognition without network access and minimizing battery usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If audio is transmitted to a remote server system for analysis and song recognition, then song identification accuracy is improved, but computational cost and time delay increase
Solution Approach 1:
The system performs preliminary music detection and continuous audio monitoring before a user explicitly requests song identification. The low-power processor continuously analyzes audio characteristics and pre-identifies music segments, so that when a user wants to know a song title, the identification is already complete or nearly complete, eliminating transmission delays.
Solution Approach 2:
The patent introduces an on-device music recognition engine as an intermediary between the microphone and remote servers. This local engine performs initial music detection and song identification, filtering out non-music audio and pre-processing music segments before any potential remote analysis, thereby reducing the need for continuous server communication and minimizing time delays.
2Ease of operation
If a specialized song identification application is launched and run, then song identification capability is improved, but user convenience deteriorates due to additional steps required
Solution Approach 1:
The system performs song identification automatically without requiring user initiation. The low-power processor continuously monitors audio, detects music presence, and identifies songs in the background. When a user wants to know a song title, the information is already available or being processed, eliminating the need for users to launch applications or press buttons.
Solution Approach 2:
The music detection and song identification processes run continuously in the background rather than being triggered on-demand. The low-power processor maintains continuous audio analysis, ensuring that song identification is always ready or nearly ready, providing immediate results when users need them without interrupting their activities.
3Measurement precision
If continuous music detection is performed by the main processor, then song identification accuracy is improved, but battery consumption increases
Solution Approach 1:
The patent divides the processing system into two segments: a low-power processor dedicated to continuous music detection and a main processor for detailed song identification. The low-power processor handles the energy-intensive continuous monitoring task using simplified algorithms, while the main processor intervenes only when music is detected or when high-precision identification is needed, significantly reducing overall battery consumption.
Solution Approach 2:
The low-power processor performs partial music detection using reduced-resolution audio analysis and simplified algorithms. It continuously monitors audio at a lower computational level, triggering full song identification only when music presence is detected, thereby achieving continuous monitoring capability with fraction of the power consumption of full main-processor-based analysis.
Data Source
AI summary
In general, the subject matter described in this disclosure can be embodied in methods, systems, and program products. A computing device stores reference song characterization data and receives digital audio data. The computing device determines whether the digital audio data represents music and then performs a different process to recognize that the digital audio data represents a particular reference song. The computing device then outputs an indication of the particular reference song.