Acoustic Echo Canceler Adaptation for Vehicle Double Talk
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
During a double talk scenario in vehicle audio systems, the acoustic echo canceler converges to a wrong solution, causing artifacts such as cancellation of desired signals, musical tones, and reverberation effects.
Innovation Solution
The audio signal is split into a parallel domain signal and an orthogonal domain signal, with separate adaptation of filter weights using constant and variable step sizes, and a total filter weight is applied to process the signal, enhancing echo cancellation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single filter weight adaptation is used for the acoustic echo canceler, then the system is simple to implement, but it converges to a wrong solution during double talk scenarios causing artifacts
Solution Approach 1:
The patent segments the filter weight adaptation into two independent domains: parallel domain and orthogonal domain. Each domain has its own filter weight (w_parallel and w_orthogonal) and step size (mu_parallel and mu_orthogonal), allowing separate optimization for each domain's specific signal characteristics and avoiding convergence to wrong solutions during double talk scenarios.
Solution Approach 2:
The patent applies different adaptation parameters (step sizes and filter weights) to different domains based on their local characteristics. The parallel domain uses parameters optimized for signals parallel to the reference, while the orthogonal domain uses parameters optimized for signals orthogonal to the reference, improving overall echo cancellation reliability.
2Productivity
If different step size values are used for parallel and orthogonal domain filter weights, then convergence is improved during double talk, but the adaptation process becomes more complex
Solution Approach 1:
The patent implements dynamic step sizes that can change based on signal conditions. The step sizes mu parallel and mu_orthogonal are updated based on the energy of signals in each domain, allowing the system to adapt the convergence speed dynamically. This enables faster convergence during double talk scenarios when needed, while maintaining stability during normal operation.
3Reliability
If the acoustic echo canceler processes signals without domain separation, then the processing is simple, but it causes cancellation of desired signals and reverberation effects
Solution Approach 1:
The patent segments the signal processing into two independent parallel paths: one for the parallel domain signal and one for the orthogonal domain signal. Each path processes signals independently using domain-specific filter weights, preventing interference between signal components and eliminating the reverberation effects that occur in single-domain processing.
Solution Approach 2:
The patent transforms the single-domain signal processing into a two-domain approach by introducing an orthogonal dimension. This dimensional separation allows the system to handle signals that are parallel to the reference and signals that are orthogonal to the reference independently, improving processing accuracy by preventing unwanted signal cancellation.
Data Source
AI summary
An example vehicle audio system includes a vehicle speaker configured to generate audio, multiple microphones, and a vehicle control module configured to receive an input signal via the multiple microphones, split the input signal into a parallel domain signal and an orthogonal domain signal, select a constant step size value for an orthogonal domain filter weight and a variable step size value for a parallel domain filter weight, adapt the orthogonal domain filter weight according to the orthogonal domain signal and the constant step size value, adapt the parallel domain filter weight according to the parallel domain signal and the variable step size value, combine the adapted orthogonal domain filter weight and the adapted parallel domain filter weight to define a total filter weight, and apply the total filter weight to the received input signal to perform a signal processing operation on the received input signal.