Cochlear Transform

Bio-inspired signal processing has emerged as a promising approach to enhance spectral decomposition and resolution in the time-frequency domain of real-life signals, drawing inspiration from the remarkable functionality of the human sensing systems, e.g., visual, auditory. In the latter, the cochlea, a nonlinear frequency analysis mechanism in the inner ear, serves as the foundation for auditory intelligence but has yet to be fully replicated in signal processing schemes. Here, we introduce the Cochlear Transform (CT), a novel bio-inspired signal processing framework that mimics the active spiral mechanics of the cochlea to produce orthogonal cochlear modes and a high resolution Time-Frequency Representation (TFR) of the input signal. CT unveils previously unseen and interpretable features not only in acoustic input signals, but also in non-acoustic ones (e.g., biological, seismic), stemming from the cochlea’s active mechanics, including acoustic distortion effects. This study is the first to demonstrate that replicating both the functional and structural aspects of the cochlea enhances TFR resolution and generalizes the approach to non-acoustic signals. CT resembles spectro-temporal receptive fields essential for auditory tasks and offers applications in data analysis, auditory machine intelligence, and bio-inspired perception. Furthermore, it opens avenues for bio-inspired feature engineering, including extensions to multivariate signals and 2D image processing.