Hearing Aid Voice Processing for Noise and Echo Cancellation

The patent divides the cancellation task into two separate adaptive filters: one for noise cancellation and one for echo cancellation. Each filter processes different aspects of the signal independently, with alternating coefficient updates. This segmentation allows each filter to specialize in one type of cancellation while avoiding the complexity of a single unified filter trying to handle both tasks simultaneously.

The patent implements periodic action through alternating updates of the two adaptive filter coefficients. During periods when the loudspeaker is active, the echo cancellation filter is updated while the noise cancellation filter maintains its coefficients. When the loudspeaker is inactive, the noise cancellation filter is updated. This periodic alternation prevents coefficient interference and maintains signal quality while managing computational complexity.

The patent employs periodic action by alternating between updating the noise cancellation filter and the echo cancellation filter based on loudspeaker activity detection. This periodic update strategy ensures that each filter is updated only when conditions are favorable, preventing coefficient interference and maintaining voice recognition accuracy while optimizing processing time through conditional execution.

The system uses feedback from loudspeaker activity detection to control the update timing of filter coefficients. The detection unit monitors whether the loudspeaker is actively outputting sound, and this feedback information determines which filter should be updated next. This feedback mechanism ensures optimal update timing, improving voice recognition accuracy while minimizing unnecessary processing operations.