Improving the Quality of Standard GMM-Based Voice Conversion Systems by Considering Physically Motivated Linear Transformations

Zorilă, Tudor-Cătălin; Erro, Daniel; Hernáez, Inmaculada

doi:10.1007/978-3-642-35292-8_4

Cited by 11 publications

(7 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A post-filter based on modulation spectrum modification is found useful to address the inherent over-smoothing issue in statistical modeling [122], such as GMM approach, which effectively compensates the global variance. The GMM approach is a parametric solution [123]- [127]. It represents a successful statistical modeling technique that works well with parallel training data.…”

Section: A Gaussian Mixture Modelsmentioning

confidence: 99%

An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning

Şişman

Yamagishi

King

et al. 2021

IEEE/ACM Trans. Audio Speech Lang. Process.

227

View full text Add to dashboard Cite

Section: A Gaussian Mixture Modelsmentioning

confidence: 99%

An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning

Şişman

Yamagishi

King

et al. 2021

IEEE/ACM Trans. Audio Speech Lang. Process.

227

View full text Add to dashboard Cite

“…Hence, the technique proposed in [16] was used in this work. This approach is based on frequency warping (FW) combined with amplitude scaling (AS), which consists in applying a linear trans-form in the cepstral domain [20]:…”

Section: Gender Conversionmentioning

confidence: 99%

Compensating Gender Variability in Query-by-Example Search on Speech Using Voice Conversion

2017

View full text Add to dashboard Cite

The huge amount of available spoken documents has raised the need for tools to perform automatic searches within large audio databases. These collections usually consist of documents with a great variability regarding speaker, language or recording channel, among others. Reducing this variability would boost the performance of query-by-example search on speech systems, especially in zero-resource systems that use acoustic features for audio representation. Hence, in this work, a technique to compensate the variability caused by speaker gender is proposed. Given a data collection composed of documents spoken by both male and female voices, every time a spoken query has to be searched, an alternative version of the query on its opposite gender is generated using voice conversion. After that, the female version of the query is used to search within documents spoken by females and vice versa. Experimental validation of the proposed strategy shows an improvement of search on speech performance caused by the reduction of gender variability.

show abstract

“…For every source-target speaker pair in the database, the FW + AS function was trained as follows. First, A was computed via (2) where W was calculated through the so-called Dynamic FW technique (see [23] for details); then, b was obtained as…”

Section: Fig 2 Parametric Definition Of An Fw Function (Left) and Anmentioning

confidence: 99%

“…In practice,

f_{a}

f_{b}

and k are set to suitable constant values and

α

is varied arbitrarily. Given W ( f ), the corresponding matrix A is calculated as

A = C \cdot W \cdot S

where S is a rectangular matrix that transforms a Mel‐cepstral vector into its corresponding discrete log‐amplitude spectrum, C is its pseudo‐inverse, and W is a sparse square matrix that captures the correspondence between original and warped spectral bins given W ( f ) [23]. As for the AS function, it can be defined using a set of weighted overlapping Hanning‐like bands, similarly as proposed in [21].…”

Section: Speaker De‐identificationmentioning

confidence: 99%

Influence of speaker de‐identification in depression detection

López-Otero

Magariños

Docío-Fernández

et al. 2017

IET signal process.

Self Cite

View full text Add to dashboard Cite

Depression is a common mental disorder that is usually addressed by outpatient treatments that favour patients' inclusion in the society. This raises the need for tools to remotely monitor the emotional state of the patients, which can be carried out via telephone or the Internet using speech processing approaches. However, these strategies lead to privacy concerns caused by the transmission of the patients' speech and its subsequent storage in servers. The use of speech deidentification to protect the privacy of these patients seems straightforward, but the influence of this procedure in the manifestation of the disease in the patients' speech has not been addressed yet. Hence, this study evaluates the performance of an automatic depression level estimation system when dealing with original and de-identified speech, in order to analyse the influence of the de-identification procedure in the detection of depression. Two de-identification approaches based on voice transformation via frequency warping and amplitude scaling are assessed, which can be applied to any speaker without additional training. Experiments carried out in the framework of the audio/visual emotion challenge 2014 show that the proposed de-identification approaches achieve promising de-identification results at the expense of a slight degradation of depression detection.

show abstract

Improving the Quality of Standard GMM-Based Voice Conversion Systems by Considering Physically Motivated Linear Transformations

Cited by 11 publications

References 10 publications

An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning

An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning

Compensating Gender Variability in Query-by-Example Search on Speech Using Voice Conversion

Influence of speaker de‐identification in depression detection

Contact Info

Product

Resources

About