Psychoacoustic sub-band filtering uses adaptive FIR coefficients and adjustable delays to lower audio processing load and memory use.
Adaptive tap calibration improves noise predictive filter convergence while cutting circuit area and power in data processing circuits.
Cumulative coefficient reordering cuts FIR hardware needs while preserving stable motor control across settling times and sampling rates.
Adjusted gradient terms across frequency bins reduce permutation errors and improve unmixing filter accuracy in convolutive audio source separation.
Adaptive multi-channel filtering tracks periodic speech components to raise signal-to-noise ratio while preserving voiced content in noise.
Heart-rate-dependent filtering extracts respiration from intracardiac pressure signals, enabling ambulatory breathing monitoring without extra sensors.
Adaptive filter coefficients extract pitch from noisy voiced speech, improving intelligibility through spectral modification and noise reduction.
Over-sampled WOLA filterbanks split physiological signals into subbands for real-time, low-power processing with robust feature extraction.
Offline transform-domain analysis updates a minimum-phase time-domain filter to reduce group delay while preserving sound quality.
Tracks voiced periodic components and mixes a delayed filtered copy with the input to improve speech intelligibility and SNR while limiting noise.
A stability factor suspends ocular artifact filtering during high-noise spikes, preserving usable EEG signals in mobile environments.
Adaptive filter coefficients reveal speech pitch more accurately in noise, improving intelligibility and perceptual quality.