Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Terashima, Ryo, Shirahata, Yuma, Kawamura, Masaya
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2507.17208
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912498055118848
author Terashima, Ryo
Shirahata, Yuma
Kawamura, Masaya
author_facet Terashima, Ryo
Shirahata, Yuma
Kawamura, Masaya
contents We present SLASH, a pitch estimation method of speech signals based on self-supervised learning (SSL). To enhance the performance of conventional SSL-based approaches that primarily depend on the relative pitch difference derived from pitch shifting, our method incorporates absolute pitch values by 1) introducing a prior pitch distribution derived from digital signal processing (DSP), and 2) optimizing absolute pitch through gradient descent with a loss between the target and differentiable DSP-derived spectrograms. To stabilize the optimization, a novel spectrogram generation method is used that skips complicated waveform generation. In addition, the aperiodic components in speech are accurately predicted through differentiable DSP, enhancing the method's applicability to speech signal processing. Experimental results showed that the proposed method outperformed both baseline DSP and SSL-based pitch estimation methods, attributed to the effective integration of SSL and DSP.
format Preprint
id arxiv_https___arxiv_org_abs_2507_17208
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SLASH: Self-Supervised Speech Pitch Estimation Leveraging DSP-derived Absolute Pitch
Terashima, Ryo
Shirahata, Yuma
Kawamura, Masaya
Audio and Speech Processing
We present SLASH, a pitch estimation method of speech signals based on self-supervised learning (SSL). To enhance the performance of conventional SSL-based approaches that primarily depend on the relative pitch difference derived from pitch shifting, our method incorporates absolute pitch values by 1) introducing a prior pitch distribution derived from digital signal processing (DSP), and 2) optimizing absolute pitch through gradient descent with a loss between the target and differentiable DSP-derived spectrograms. To stabilize the optimization, a novel spectrogram generation method is used that skips complicated waveform generation. In addition, the aperiodic components in speech are accurately predicted through differentiable DSP, enhancing the method's applicability to speech signal processing. Experimental results showed that the proposed method outperformed both baseline DSP and SSL-based pitch estimation methods, attributed to the effective integration of SSL and DSP.
title SLASH: Self-Supervised Speech Pitch Estimation Leveraging DSP-derived Absolute Pitch
topic Audio and Speech Processing
url https://arxiv.org/abs/2507.17208