Quantile-Frequency Analysis and Deep Learning for Signal Classification

This paper proposes a new method for signal classification based on combination of deep-learning (DL) image classifiers and recently introduced nonlinear spectral analysis technique called quantile-frequency (QFA). The QFA converts one-dimensional into two-dimensional representation quantile periodograms (QPER) which represent the signal’s oscillatory behavior in frequency domain at different quantiles. With moving window, can also covert sequence such representations, short-time periodograms, that are localized time to time-dependent or nonstationary properties. DL take these representations as input classification. benefit this QFA-DL comparison with traditional frequency-domain power spectrum spectrogram is demonstrated by numerical experiment using real-world ultrasound signals from nondestructive evaluation application.