Sound Collection Apparatus Using Image-Guided Beamforming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional beam forming technologies for voice recognition struggle to precisely extract target sounds in noisy environments, often canceling target sounds when nearby sounds are similar to noise, leading to reduced extraction precision and ineffective noise cancellation.
Innovation Solution
A sound collection apparatus and method that uses image capturing to determine sound collection and noise directions, separating sounds using beam forming and spectral subtraction to accurately isolate and cancel noise from the target sound.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If beam forming technology is used to separate target sound in designated direction, then sound source separation is achieved, but target sound extraction precision drops in noisy environments
Solution Approach 1:
The patent segments the sound field into multiple directional components using beam forming, separating the target sound from environmental noise by spatial segmentation. The microphone array divides the acoustic space into different directional beams, allowing independent processing of target and noise components.
Solution Approach 2:
The patent introduces an intermediary variable (attention score) that measures the similarity between noise sound and target sound. This intermediary enables the system to intelligently control the noise canceling process, applying spectral subtraction only when the attention score exceeds a threshold, thus preventing target sound cancellation.
2Object-affected harmful factors
If noise sound is canceled using spectral subtraction method, then noise reduction is achieved, but target sound is also canceled when similar sounds are present
Solution Approach 1:
The patent implements a feedback mechanism where the attention score (measuring similarity between noise and target sound) controls the spectral subtraction process. When the attention score is high, spectral subtraction is suppressed to prevent target sound cancellation; when low, noise canceling is applied. This feedback loop dynamically adjusts noise canceling based on real-time sound analysis.
Solution Approach 2:
The system dynamically adjusts the noise canceling intensity based on the calculated attention score. The spectral subtraction coefficient varies dynamically according to the similarity between noise and target sound, allowing the system to adapt to changing acoustic conditions and prevent over-canceling of target sound.
3Object-affected harmful factors
If sound in opposite direction is used for noise canceling, then noise canceling effect is achieved when noise source exists, but conversation voice is canceled when human is present in opposite direction
Solution Approach 1:
The patent uses the attention score as an intermediary to determine whether sound from the opposite direction should be used for noise canceling. The attention score quantifies the similarity between the opposite-direction sound and the target sound, enabling intelligent decision-making about noise canceling application.
Solution Approach 2:
The system changes the spectral subtraction coefficient parameter based on the attention score. When the attention score exceeds a threshold, the coefficient is reduced or set to zero, preventing target sound cancellation. This parameter adjustment allows flexible control of noise canceling effectiveness.
4Ease of operation
If conventional noise canceling is applied without checking sound similarity, then processing simplicity is maintained, but extraction precision drops when noise resembles target sound
Solution Approach 1:
The patent performs preliminary analysis by calculating the attention score (similarity between noise and target sound) before applying spectral subtraction. This preliminary check ensures that noise canceling is only applied when appropriate, preventing target sound cancellation while maintaining processing efficiency through a simple threshold-based decision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables more precise extraction of target sounds by distinguishing between sound collection and noise directions, improving voice recognition accuracy even in complex acoustic environments.
Implementation Method 1
there is a beam forming technology in which a sound source in a designated direction is separated using a microphone array equipped with a plurality of microphones
Implementation Method 2
the noise sound is canceled from the separated target sound using a spectral subtraction method or the like to precisely extract the target sound
Data Source
AI summary
A sound collection direction is decided based upon an area of an object in a captured image obtained by image capturing of a periphery and a sound collection target position input as a position of a sound collection target. A noise direction is decided based upon an arrangement of the object in the captured image. A sound collected from the periphery is separated into a sound in the sound collection direction and a sound in the noise direction, and noise canceling on the sound in the sound collection direction is performed using the sound in the noise direction.