Glottal Flow Synthesis for Whisper-to-Speech Conversion

Abstract

Whisper-to-speech conversion is motivated by laryngeal disorders, in which malfunction of the vocal folds leads to loss of voicing. Many patients with laryngeal disorders can still produce functional whispers, since these are characterised by the absence of vocal fold vibration. Whispers therefore constitute a common ground for speech rehabilitation across many kinds of laryngeal disorder. Whisper-to-speech conversion involves recreating natural-sounding speech from recorded whispers, and is a non-invasive and non-surgical rehabilitation that can maintain a natural method of speaking, unlike the existing methods of rehabilitation. This paper proposes a new rule-based method for whisper-to-speech conversion that replaces the noisy whisper sound source with a synthesised speech-like harmonic source, while maintaining the vocal tract component unaltered. In particular, a novel glottal source generator is developed in which whisper information is used to parameterise the excitation through a high-quality glottis model. Evaluation of the system against the standard pulse train excitation method reveals significantly improved performance. Since our method is glottis-based, it is potentially compatible with the many existing vocal tract component adaptation systems.

The Detection of Parkinson's Disease From Speech Using Voice Source Information

“…The spectral tilt of the glottal source is computed over several glottal periods (including both glottal open and closed phases), and therefore the IAIF method does not call for the extraction of glottal closure instants (GCIs). During the past two decades, IAIF has been used in many areas, such as in parametric speech synthesis ([52] [53] [54] [55]), speaking style conversion [56], the detection of stress [57], and depression [58], as well as in emotion recognition [59] [60]. For a detailed description of the IAIF method, the reader is referred to Section II-B in the work of Raitio et al [52].…”

Automated Assessment of Glottal Dysfunction Through Unified Acoustic Voice Analysis

“…In the presence of noise in the glottis signal, we showed that F GF captures the position of the dominant frequency region of this noise. 21 Therefore, in the control group we observe an F GF distribution around a few hundred Hertz, the order of magnitude of vocal fold vibration. The increase of F GF with the degree of impairment follows the increasing amount of noise in the glottal signal, linked to the progressive loss of phonation.…”

“…Note that the GFM-IAIF glottis filter is fully causal compared to Equation 1, yet it does not affect the magnitude spectrum, from which are extracted the spectral parameters (eg, F GF , B GF , and F ST ) that we now use for analysis of dysphonic speech. GFM-IAIF has also recently been demonstrated in the conversion of postlaryngectomy speech to phonated speech, 21 lending empirical support to its effectiveness at modeling dysfunctional speech.…”

“…29 Close correlation between vocal effort and spectral tilt 30,31 was a major motivation for using a threepole GFM. This has been validated for voice quality analysis and modification, 32 expressive singing and speech synthesis, 33,34 whisper-to-speech conversion, 21 and in this paper is now considered for analyzing the effects of dysfunction on glottis parameters. For convenience, we will use the term "glottal formant" even for speakers who lack a glottis.…”

“…22 It has recently been used to reconstruct a speech glottal source from analyses of whisper source, using a single decomposition model. 21 This successful extrapolation to non-speech voice sources has led − in this paper − to its application to a set of voice recordings obtained from patients who have undergone glottal or larynx treatment and surgery (including excision), alongside baseline normal speech. The analysis will demonstrate the ability of GFM-IAIF to derive parameters from voice source signals that differentiate the category of impairment present.…”