Cochlear transform: A novel bio-inspired digital stethoscope for biomedical signals

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 human sensory systems such as vision and audition. In the auditory system, the cochlea provides frequency-selective analysis through its organized tonotopic structure, which has inspired numerous signal processing models. However, existing approaches are typically formulated as filterbanks or hardware implementations and do not provide a unified transform framework that supports both decomposition and reconstruction. In this work, we introduce the Cochlear Transform (CT), a bio-inspired signal processing framework that incorporates cochlear-inspired frequency organization into a continuous transform domain. The proposed CT produces structured time–frequency representations and enables adaptive signal decomposition with improved interpretability. The resulting representation reveals fine-grained signal features in both acoustic and non-acoustic signals (e.g., biological and seismic), demonstrating the generality of the framework. Quantitative and qualitative evaluations show that CT provides competitive performance in terms of time–frequency concentration, mode separation, and noise robustness compared to established methods such as EWT and VMD. Furthermore, the transform exhibits stable reconstruction behavior and supports analysis across diverse signal types. The proposed framework provides a structured and interpretable representation that may be useful for data analysis, auditory-inspired machine intelligence, and bio-inspired feature engineering, with potential extensions to multivariate signals and image processing.