Harmonic-Aligned Frame Mask Based on Non-Stationary Gabor Transform with Application to Content-Dependent Speaker Comparison

Abstract

We propose harmonic-aligned frame mask for speech signals using non-stationary Gabor transform (NSGT). A frame mask operates on the transfer coefficients of a signal and consequently converts the signal into a counterpart signal. It depicts the difference between the two signals. In preceding studies, frame masks based on regular Gabor transform were applied to single-note instrumental sound analysis. This study extends the frame mask approach to speech signals. For voiced speech, the fundamental frequency is usually changing consecutively over time. We employ NSGT with pitch-dependent and therefore time-varying frequency resolution to attain harmonic alignment in the transform domain and hence yield harmonic-aligned frame masks for speech signals. We propose to apply the harmonic-aligned frame mask to content-dependent speaker comparison. Frame masks, computed from voiced signals of a same vowel but from different speakers, were utilized as similarity measures to compare and distinguish the speaker identities (SID). Results obtained with deep neural networks demonstrate that the proposed frame mask is valid in representing speaker characteristics and shows a potential for SID applications in limited data scenarios. Index Terms: Non-stationary Gabor transform, frame mask, harmonic alignment, pitchdependent frequency resolution, speaker feature, speaker comparison 1 Introduction Time-frequency (TF) analysis is the foundation of audio and speech signal processing. The shorttime Fourier transform (STFT) is a widely used tool, which can be effectively implemented by FFT [1]. STFT features straightforward interpretation of a signal. It provides uniform time and frequency resolution with linearly-spaced TF bins. The corresponding theory was generalized in the framework of Gabor analysis and Gabor frames [2, 3, 4].Signal synthesis is an important application area of time-frequency transforms. Signal modification, denoising, separation and so on can be achieved by manipulating the analysis coefficients to synthesize a desired one. The theory of Gabor multiplier [5] or, in general terms, frame multiplier [6,7] provides a basis for the stability and invertibility of such operations. A frame multiplier is an

Application of frame multipliers for the extraction of curve squeals from train signals

“…[15], welche dann in der Akustik eingesetzt werden, etwa für die Signaltrennung [27]. Rahmen-Multiplikatoren treten daher nicht nur als mathematische Objekte auf [6,25], es werden auch deren Anwendungen studiert, wie zum Beispiel für die Entfernung von perzeptiv irrelevanten Zeit-Frequenz-Punkten [5], oder auch als Möglichkeit, die Ähnlichkeit von Instrumenten [17] oder Sprechern [14] zu beurteilen. Ein einfaches Beispiel, das etwa in der computergestützten auditorischen Szenenanalyse (CASA [28]) aber auch für das Beispiel in Abschn.…”