Echo Suppression Using Multi-Delay Adaptive Filter Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing echo suppression technologies fail to secure sufficient echo suppression due to erroneous learning of adaptive filter coefficients, leading to inefficiencies and increased processing loads.
Innovation Solution
An echo suppressing device that generates multiple delay reference signals, obtains convergence values of an adaptive filter, and corrects the filter coefficients to minimize changes with delay time, thereby reducing echo removal errors and processing loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional echo suppression using adaptive filter coefficients is applied, then echo removal is attempted, but erroneous learning of filter coefficients causes insufficient echo suppression
Solution Approach 1:
The patent applies preliminary action by performing filter coefficient learning in advance before actual echo suppression operation. Multiple adaptive filters are pre-trained using different delay times, and the most appropriate filter is selected beforehand. This preliminary learning process ensures that when echo suppression is needed, the system uses pre-validated filter coefficients rather than attempting real-time learning, thereby avoiding erroneous learning and ensuring reliable echo suppression.
2Reliability
If multiple delay reference signals are generated and multiple adaptive filters are trained, then echo suppression reliability improves, but processing load increases
Solution Approach 1:
The patent performs all adaptive filter training and convergence value calculation in advance as preliminary actions. Multiple delay reference signals are generated and processed beforehand to obtain convergence values, which are then stored for later use. During actual echo suppression operation, the system simply selects the pre-computed convergence values and applies the corresponding filter coefficients, avoiding the heavy computational load of real-time training while maintaining high reliability.
Solution Approach 2:
The patent creates multiple copies of adaptive filters, each trained with different delay times. Instead of training one filter repeatedly under different conditions during operation, the system pre-creates multiple filter copies with different delay characteristics. During operation, the appropriate copy is selected based on the actual delay time, avoiding redundant training computations and reducing processing load while maintaining comprehensive echo suppression capability.
3Adaptability or versatility
If filter coefficients are continuously updated during operation, then adaptation to changing acoustic paths is improved, but erroneous learning occurs and echo suppression deteriorates
Solution Approach 1:
The patent performs all filter coefficient updates and adaptations in advance during the learning phase rather than continuously during operation. Multiple adaptive filters are pre-trained with different delay times and acoustic path conditions before actual use. During operation, the system selects the pre-adapted filter that best matches the current acoustic path, avoiding continuous updates that could cause erroneous learning and maintain reliable echo suppression performance.
Data Source
AI summary
An echo suppressing device removes an echo from an input signal picked up by a microphone of a terminal including a speaker and the microphone. Included are a reference signal adjustment unit that generates a plurality of delay reference signals by applying a plurality of delay times different from one another to a reference signal transmitted through a receiver-side signal path for transmitting a signal to the speaker; a filter generation unit that obtains a plurality of convergence values of an adaptive filter based on each of the plurality of delay reference signals; a filter correction unit that compares the plurality of convergence values and makes a first filter coefficient smaller than the convergence values, the first filter coefficient being a filter coefficient for an element that does not change in accordance with a change in the delay times; and an echo removal unit that removes a linear echo.