Acoustic Object Extraction via Spectral Similarity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting acoustic objects using multiple beamformers are not comprehensive and struggle to effectively distinguish and separate target acoustic objects from noise and other non-target sounds, leading to suboptimal extraction performance.
Innovation Solution
The proposed acoustic object extraction apparatus employs multiple microphone arrays and beamforming processors to generate signals, which are then processed by a common component extractor that divides spectra into subbands, calculates similarity degrees, and applies spectral gains to extract and isolate acoustic objects based on spectral shape differences, enhancing extraction performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple beamformers are used to extract acoustic objects, then the extraction capability is enhanced, but the ability to distinguish target sounds from noise and non-target sounds deteriorates
Solution Approach 1:
The spectrum is divided into multiple subbands, allowing the system to analyze and compare spectral shapes at different frequency resolutions. This segmentation enables better distinction between target acoustic objects and noise by examining spectral characteristics across multiple frequency bands rather than as a whole
Solution Approach 2:
The patent introduces a new dimension of spectral shape comparison in the frequency domain by calculating similarity degrees based on spectral shapes. This adds a spectral dimension to the extraction process, complementing the spatial dimension provided by multiple beamformers, and enables more accurate target identification
2Reliability
If spectral shapes are similar between target and non-target sounds, then extraction becomes difficult, but using conventional methods still results in noise contamination
Solution Approach 1:
The patent applies spectral gain to only those frequency components where the spectral shape similarity degree exceeds a predetermined threshold. This partial action approach allows the system to selectively enhance target frequencies while suppressing noise frequencies, rather than applying uniform processing across all frequencies
Solution Approach 2:
The system dynamically adjusts spectral gains based on calculated similarity degrees between spectral shapes. By changing the gain parameter adaptively according to spectral similarity, the system can enhance target sounds with similar spectral characteristics while suppressing noise, resolving the contradiction between reliable extraction and noise contamination
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In the acoustic object extraction device (100), beam forming processing units generate a first acoustic signal by beam forming in an arrival direction of a signal from an acoustic object with respect to a microphone array and generate a second acoustic signal by beam forming in an arrival direction of a signal from the acoustic object with respect to a microphone array, and a common component extraction unit extracts, on the basis of a similarity between the spectrum of the first acoustic signal and the spectrum of the second acoustic signal and from the first acoustic signal and the second acoustic signal, a signal containing a common component corresponding to the acoustic object. The common component extraction unit divides the spectrums of the first acoustic signal and the second acoustic signal into a plurality of frequency sections and calculates a similarity for each of the frequency sections.