Adaptive Differential Microphone Speech Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech detectors with closely-spaced microphones struggle to adapt to varying microphone configurations and user handling orientations, leading to poor performance in detecting speech that does not originate from the forward direction, and require calibration of microphone power levels.
Innovation Solution
A method and system using an adaptive differential microphone (ADM) that normalizes signals from multiple microphones, constructs directional nulls, and computes ratios to detect speech by steering the null towards the target source, eliminating the need for precise microphone calibration and adapting to different orientations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional speech detectors assume speech is louder on one microphone than the other to use inter-microphone level differences, then speech detection can be performed, but the positioning of microphones is constrained and performance deteriorates when speech does not originate from the forward direction
Solution Approach 1:
The patent applies dynamics by making the null direction of the adaptive differential microphone adjustable and reconfigurable. The system dynamically adapts the beamformer weights to steer the null towards the actual speech source direction, rather than assuming a fixed forward direction. This allows the speech detector to maintain high accuracy regardless of the speaker's position relative to the device orientation.
Solution Approach 2:
The patent changes the parameter of microphone signal weighting by applying adaptive beamforming techniques. The system adjusts the weights applied to signals from different microphones based on the estimated speech source direction, enabling the detector to adapt to various orientations and positions of the speech source without requiring fixed microphone positioning assumptions.
2Reliability
If speech enhancement algorithms use a generalised sidelobe canceller with beamformer and adaptive filter, then speech enhancement can be achieved, but poor speech detector performance leads to suppression of target signal and reinforcement of interfering sources
Solution Approach 1:
The patent implements feedback by using the output of the adaptive differential microphone to control the adaptation of both the beamformer and the second-stage adaptive filter. The speech detector continuously monitors the signal and adjusts the beamformer weights accordingly, creating a closed-loop system where detection accuracy directly controls the enhancement process, preventing suppression of target speech and reinforcement of interfering sources.
3Volume of moving object
If microphones are closely-spaced to one another, then the device size can be reduced, but the speech detector must rely on inter-microphone level differences which constrains microphone positioning
Solution Approach 1:
The patent applies universality by creating a speech detection system that works effectively with closely-spaced microphones regardless of their specific arrangement or orientation. The adaptive differential microphone structure with steerable null provides a universal solution that accommodates various microphone topologies and device orientations, making the system versatile for different form factors and applications without requiring specific microphone positioning.
Data Source
AI summary
A method for detecting speech using a first microphone adapted to produce a first signal (x), and a second microphone adapted to produce a second signal (x2), the method comprising the steps of:(i) applying gain to the second signal to produce a normalised second signal, which signal is normalised relative to the first signal;(ii) constructing one or more signal components from the first signal and the normalised second signal;(iii) constructing an adaptive differential microphone (ADM) having a constructed microphone response constructed from the one or more signal components which response has at least one directional null;(iv) producing one or more ADM outputs (yf, yb) from the constructed microphone response in response to detected sound;(v) computing a ratio of a parameter of either a first signal component or a constructed microphone response to a parameter of an output of the ADM;(vi) comparing the ratio to an adaptive threshold value;(vii) detecting speech if the ratio is greater than or equal to the adaptive threshold value.