Faster exact invertible audio spectral transform saves memory and time
Exact Factorisation and Fast Computation of Invertible Constant-Q Transforms
Sound
Summary
The constant-Q transform (CQT) helps analyze sounds using frequencies spaced like musical notes. The authors found a new way to calculate this transform on GPUs that makes it much faster and uses less memory, while still perfectly reconstructing the original sound. Their method, called Flash-CQT, rearranges and simplifies the math steps to speed up the process. It works well on different GPU models and keeps errors extremely low, making it suitable for audio analysis and machine learning with sound.
What this means in practice
- •For audio software developers: Create faster, memory-efficient audio spectral analysis tools with exact invertibility using Flash-CQT on GPUs.
- •For machine learning engineers in audio: Use Flash-CQT as a practical front end for training neural networks with precise spectral features and faster data processing.
Authors
Facundo Franchino, Eloi Moliner, Vesa Välimäki
Abstract
The constant-Q transform (CQT) represents audio on a logarithmic frequency axis. Its nonstationary Gabor formulation is exactly invertible, but the unequal numbers of time coefficients in its bands complicate GPU computation. An exact factorisation combines spectral selection, conjugation, windowing, and reordering into a fixed map between one packed Fourier transform and the shorter band inverse transforms. The factors give waveform reconstruction, real adjoints for backpropagation, and bounds on arithmetic depth and block width; overlapping slices permit streaming with bounded memory. Tests on two GPU models show that Flash-CQT reduces analysis-synthesis round-trip time by factors of two to eight relative to a baseline computing the same CQT. The proposed implementation also uses over 30% less peak temporary workspace and reaches a negligible reconstruction error, with a signal-to-noise ratio of about 130 dB, in single-precision floating-point arithmetic. These advances make Flash-CQT a practical, computationally efficient front end for spectral analysis and modern audio machine-learning systems.