This paper describes a novel algorithm for recovering time-domain signal from quantized amplitude and phase spectrogram, which is applicable for spectrogram-based audio coding. In order to obtain a better quality sound, a phase reconstruction technique is first applied with constraint for keeping phase in each time-frequency bin within each quantization range, and then, time-domain signal is re...

Frequency-resolved optical gating (FROG) is a technique to measure ultrashort laser pulses that optically constructs a spectrogram of a laser pulse. A two-dimensional (2-D) phase retrieval algorithm is used to extract the intensity and phase of a pulse from its spectrogram. We have improved a recently presented principal component generalized projections algorithm (PCGPA) making it easier to im...

We apply the modal distribution, a high-resolution time-frequency distribution, to the study of sung musical passages. Evidence is presented comparing the modal distribution with the spectrogram for a set of synthetic signals which emulate human singing. We then compare the two techniques for sung passages from four student sopranos with respect to measures of the instantaneous frequency and am...

We study the problem of separating audio sources from a single linear mixture. The goal is to find a decomposition of the single channel spectrogram into a sum of individual contributions associated to a certain number of sources. In this paper, we consider an informed source separation problem in which the input spectrogram is partly annotated. We propose a convex formulation that relies on a ...

In this paper, the usage of pseudo 2-dimensional Hidden Markov Models for speech recognition is discussed. This image processing method should better model the timefrequency structure in speech signals. The method calculates the emission probability of a standard HMM by embedded HMMs for each state. If a temporal sequence of spectral vectors is imagined as a spectrogram, this leads to a 2-dimen...

In this paper we describe a method of audio chord estimation than does not rely on any machine learning technique. We calculate a beat-synchronized spectrogram with high time and frequency resolution. The sequence of chroma vectors (CRP features based on constant-Q transform) obtained from spectrogram is smoothed using self-similarity matrix before the actual chord recognition. Binary chord tem...

In this course project I investigated machine learning approaches on separating speech signals from background noise. Keywords—MFCC, SVM, noise separation, source separation, spectrogram