Nonlinear Acoustic Echo Cancellation Using Back-EMF Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hands-free acoustic telecommunication systems face challenges in eliminating nonlinear echo components caused by loudspeaker nonlinearities, which conventional linear adaptive filters cannot effectively address, leading to suboptimal audio quality and full-duplex functionality issues.
Innovation Solution
A Nonlinear Acoustic Echo Canceller (NL-AEC) that models the loudspeaker as a Hammerstein system using the Back Electromotive Force (BEMF) signal, allowing for the estimation and compensation of nonlinear echo components, thereby improving echo cancellation performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If linear adaptive filters are used in acoustic echo cancellers, then device complexity is reduced and ease of manufacture is improved, but nonlinear echo components caused by loudspeaker nonlinearities cannot be correctly estimated and suppressed
Solution Approach 1:
The patent changes the fundamental parameter of the echo canceller from linear to nonlinear processing. Specifically, it employs nonlinear adaptive filters (such as Volterra filters, neural networks, or NARMA models) that can capture and compensate for the nonlinear distortions introduced by loudspeakers driven into their nonlinear range, thereby correctly estimating and suppressing nonlinear echo components while maintaining full-duplex operation
Solution Approach 2:
The patent introduces dynamic nonlinear modeling that adapts to the changing operating conditions of the loudspeaker. By using adaptive nonlinear filters that continuously adjust their parameters based on the input signal characteristics, the system can track and compensate for nonlinear echo components across different sound pressure levels and operating states, improving reliability without requiring complex manual calibration
2Power
If loudspeaker is driven into nonlinear range to achieve adequate sound pressure levels, then audio output capability is improved, but nonlinearly distorted echo components are contributed to the near-end microphone signal
Solution Approach 1:
The patent converts the harmful nonlinear distortion into a useful signal by using the Back Electromotive Force (BEMF) generated by the loudspeaker coil as an additional input to the echo canceller. The BEMF contains information about the loudspeaker's mechanical motion and nonlinearities, which is used to create a more accurate model of the nonlinear echo path, thereby enabling effective cancellation of the distorted echo components
Solution Approach 2:
The patent introduces the BEMF signal as an intermediary that bridges the electrical and mechanical domains. This intermediate signal provides insight into the loudspeaker's nonlinear behavior and is used to generate compensation signals that cancel the nonlinear echo components in the microphone signal, effectively mediating between the loudspeaker output and the echo cancellation process
3Device complexity
If conventional linear acoustic echo cancellers are used, then device complexity is kept low, but full-duplex functionality deteriorates due to inability to suppress nonlinear echo components
Solution Approach 1:
The patent changes the mathematical model from linear to nonlinear, enabling the echo canceller to handle the complex nonlinear relationships in the acoustic path. This allows full-duplex operation where both near-end and far-end speakers can talk simultaneously without echo interference, as the nonlinear filter can accurately predict and subtract nonlinear echo components from the microphone signal
Solution Approach 2:
The patent creates a universal echo cancellation system that handles both linear and nonlinear echo components through a unified nonlinear filtering approach. The system can adapt to different operating conditions, loudspeaker types, and acoustic environments, providing robust full-duplex functionality across diverse applications without requiring separate processing paths
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
The NL-AEC effectively reduces nonlinear echo components, enhancing audio quality and maintaining full-duplex functionality by employing the BEMF signal to model the loudspeaker's nonlinearities, which conventional methods struggle with.
Implementation Method 1
The method employs the Back Electro-Motive Force (BEMF) signal induced in the loudspeaker coil due to its movement in a static magnetic field to estimate the position and/or velocity of the coil
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method and system for determining and compensating for a non-linearity in a hands-free acoustic telecommunication device is disclosed. The method determines a back electromotive force signal induced in a loudspeaker from at least one of a coil voltage, a current signal and estimates of coil resistance and inductance, estimating at least one of a cone position and a cone velocity from the BEMF integrated with respect to time and determining an estimate of an echo value from a series connection of an estimated inverse of a force factor function primitive and the estimated acoustic impulse response; and outputting the estimated echo value.