Audio Scene Selection Using Acoustic Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audio signal processing systems face challenges in selecting relevant audio sources from multiple recordings of the same event, particularly when devices are in close proximity, as they often rely on basic criteria like 'nearest' or 'loudest', missing subtle sound qualities recorded from the periphery.
Innovation Solution
An apparatus and method that classify and select audio signals by determining dominant audio sources based on audio events, signal directions, and time intervals, using processors and memory to analyze and combine signals from multiple devices, ensuring accurate selection of relevant audio sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If basic selection criteria (nearest, loudest) are used to select audio sources, then the selection process is simple and fast, but subtle sound qualities from peripheral sources are missed
Solution Approach 1:
The system changes the selection parameters from basic criteria (distance, volume) to acoustic scene analysis parameters (reverberation characteristics, direct-to-reverberant energy ratio, spectral content). This allows peripheral sources with subtle sound qualities to be identified and selected based on their unique acoustic signatures rather than being discarded by simple proximity or volume thresholds.
Solution Approach 2:
The patent replaces mechanical selection criteria (physical proximity, signal amplitude) with acoustic analysis mechanisms (reverberation pattern recognition, scene classification). This substitution enables the system to detect and preserve subtle sound qualities that mechanical criteria would filter out, while maintaining automated operation.
2Loss of information
If multiple audio sources from different positions are recorded, then more comprehensive audio coverage is achieved, but selecting the relevant audio source becomes difficult
Solution Approach 1:
Each audio source automatically provides acoustic scene information (reverberation characteristics, directional cues, spectral content) that serves as its own identification signature. The system uses these self-provided characteristics to automatically classify and select sources without requiring complex external analysis or manual intervention, thus managing complexity while maintaining comprehensive coverage.
Solution Approach 2:
The system performs preliminary acoustic scene analysis on all recorded sources, classifying them by their acoustic characteristics before selection is needed. This pre-processing organizes the multiple sources into categorized groups, making the subsequent selection process simpler while preserving comprehensive audio coverage from all classified sources.
3Loss of information
If audio sources from peripheral positions are included, then subtle sound qualities are captured, but the selection process becomes more complex
Solution Approach 1:
The system introduces acoustic scene parameters (reverberation time, spectral density, directional information) as selection criteria, enabling peripheral sources with subtle qualities to be identified. These parameters provide an automated classification mechanism that manages complexity by objectively categorizing sources based on measurable acoustic properties rather than requiring complex subjective evaluation.
Data Source
AI summary
An apparatus comprising: an audio analyzer configured to determine for a set of received audio signals at least one dominant audio signal source; and a selector configured to select from the set of audio signals at least one audio signal dependent on the at least one dominant audio signal source.