Time–frequency analysis

About: Time–frequency analysis is a research topic. Over the lifetime, 5407 publications have been published within this topic receiving 104346 citations.

Time-Frequency Jigsaw Puzzle: Adaptive multiwindow and multilayered Gabor expansions

Abstract: We describe a new adaptive multiwindow Gabor expansion, which dynamically adapts the windows to the signal's features in time-frequency space. The adaptation is based on local time-frequency sparsity criteria, and also yields as by-product an expansion of the signal into layers corresponding to different windows. As an illustration, we show that simply using two different windows with different sizes leads to decompositions of audio signals into transient and tonal layers. We also discuss potential applications to transient detection and denoising.

Analysis of Modulated Multivariate Oscillations

Abstract: The concept of a common modulated oscillation spanning multiple time series is formalized, a method for the recovery of such a signal from potentially noisy observations is proposed, and the time-varying bias properties of the recovery method are derived. The method, an extension of wavelet ridge analysis to the multivariate case, identifies the common oscillation by seeking, at each point in time, a frequency for which a bandpassed version of the signal obtains a local maximum in power. The lowest-order bias is shown to involve a quantity, termed the instantaneous curvature, which measures the strength of local quadratic modulation of the signal after demodulation by the common oscillation frequency. The bias can be made to be small if the analysis filter, or wavelet, can be chosen such that the signal's instantaneous curvature changes little over the filter time scale. An application is presented to the detection of vortex motions in a set of freely drifting oceanographic instruments tracking the ocean currents.

Phase-Aware Speech Enhancement Based on Deep Neural Networks

Abstract: Short-time frequency transform (STFT) is fundamental in speech processing Because of the difficulty of processing highly unstructured STFT phase, most speech-processing algorithms only operate with STFT magnitude, leaving the STFT phase far from explored However, with the recent development of deep neural network (DNN) based speech processing, eg, speech enhancement and recognition, phase processing is becoming more important than ever before as a new growing point of DNN-based methods In this paper, we propose a phase-aware speech enhancement algorithm based on DNN Specifically, in the training stage, when incorporating phase as a target, our core idea is to transform an unstructured phase spectrogram to its derivative along the time axis, ie, instantaneous frequency deviation (IFD), which has a similar structure with its corresponding magnitude spectrogram We further propose to optimize both IFD and magnitude jointly in a multiobjective learning framework In the test stage, we propose a postprocessing method to recover the phase spectrogram from the estimated IFD Experimental results demonstrate the effectiveness of the proposed method