Adaptive Extended TDAC Audio Processing System for Improved Coding Efficiency

Researchers at Stanford have developed an adaptive audio filter bank system that aims to improve the efficiency and quality of perceptual audio coding by analyzing and synthesizing audio more precisely and dynamically adapting to the audio signal characteristics.

Modern digital audio systems rely on perceptual audio codecs to efficiently store and transmit high-quality audio. These codecs use filter banks to convert audio from the time domain to the frequency domain for data compression. Time-domain aliasing cancellation (TDAC) transforms are a widely used filter bank approach, but their frequency and time resolution can be limited. Attempts to improve frequency resolution have affected how flexibly and rapidly they can traverse the time–frequency resolution trade-off. Both filterbank resolution and rapid time-frequency adaptivity are essential heuristics for audio coding quality.

This system introduces an adaptive extended TDAC transform that dynamically switches between different transform configurations based on the characteristics of the input audio. Optimized extended windows enhance the transform's stopband attenuation, and the audio coder transitions rapidly and without distortion between configurations.. The result is the potential for superior coding quality across a wider range of audio content, with applications in music streaming, broadcast audio, and telecommunications.

Perceptual Audio Coding Schematic (Image courtesy Ryan Wixen and Marina Bosi)

Stage of Development - Proof of concept