Voice Conversion

— Limited functionality of face detection algorithm are still facing problems in real environment. This problem is marked as Sparse Fingerprint Classification Algorithm (SFCA). There are two phases training and testing. Parts of images are obtained in training phase and in testing phase every part of image is converted into binary sparse format. The binary representation marks for their respective categories and maximum marked category decides the identity of the image. SFCA works efficient when the size of data file is limited.

The goal of our research work was the application of Voice Conversion on high-quality speech. Our main interests were established in the improvement of current systems quality and the use of high-quality speech. To achieve this, we focused our motivation into the study of improved spectral envelope modeling and timbre modification.

Learners of a second language practice their pronunciation by listening to and imitating utterances from native speakers. Recent research has shown that choosing a well-matched native speaker to imitate can have a positive impact on pronunciation training. Here we propose a voice-transformation technique that can be used to generate the (arguably) ideal voice to imitate: the own voice of the learner with a native accent. Our work extends previous research, which suggests that providing learners with prosodically corrected versions of their utterances can be a suitable form of feedback in computer assisted pronunciation training. Our technique provides a conversion of both prosodic and segmental characteristics by means of a pitch-synchronous decomposition of speech into glottal excitation and spectral envelope. We apply the technique to a corpus containing parallel recordings of foreign-accented and native-accented utterances, and validate the resulting accent conversions through a series of perceptual experiments. Our results indicate that the technique can reduce foreign accentedness without significantly altering the voice quality properties of the foreign speaker. Finally, we propose a pedagogical strategy for integrating accent conversion as a form of behavioral shaping in computer assisted pronunciation training.

Abstract—The crosslingual voice conversion problem refers to the replacement of a speaker’s timbre or vocal identity in a recorded sentence, assuming that the source speaker and target speaker use different languages. This problem differs from typical voice conversion in the sense that the mapping of acoustical features cannot depend on time-aligned recordings of source and target speakers uttering the same sentences. This paper presents an overview of a general crosslingual voice conversion system and discusses the most important techniques used in each step of the conversion process. Keywords-Crosslingual Voice Conversion; Timbre transformation; Vocal Identity;

This work addresses a study of the GMM-based approach to achieve full-quality speaker timbre conversion. In general, high-quality voice conversion requires accurate spectral enve- lope estimates, resulting in high-dimensional feature vectors and relatively high computational. Aiming to achieve low- dimensional processing, accurate envelope estimates of the speakers are mel-frequency scaled and projected onto the space defined by a subset of the principal components. The GMM- based features conversion is then performed in the reduced space. Our experimental findings confirm that this strategy pro- vides benefits, especially observed on the resulting converted speech quality, with a significant computational cost reduction.

In voice conversion, frame-level mean and variance normalization is typically used for fundamental frequency (F0) transformation, which is text-independent and requires no parallel training data. Some advanced methods transform pitch contours instead, but require either parallel training data or syllabic annotations. We propose a method which retains the simplicity and text-independence of the frame-level conversion while yielding high-quality conversion. We achieve these goals by (1) introducing a text-independent tri-frame alignment method, (2) including delta features of F0 into Gaussian mixture model (GMM) conversion and (3) reducing the well-known GMM oversmoothing effect by F0 histogram equalization. Our objective and subjective experiments on the CMU Arctic corpus indicate improvements over both the mean/variance normalization and the baseline GMM conversion.

Abstract—The goal of this project is to convert a given speaker's speech (the Source speaker) into another identified voice (the Target speaker) as well as analysing the face animation of the source to animate a 3D avatar imitating the source facial movements. We assume we have at our disposal a large amount of speech samples from the source and target voices with a reasonable amount of parallel data. Speech and video are processed separately and recombined at the end. Voice conversion is obtained in two steps: a voice ...

In this work we present an approach to perform voice timbre con- version from unpaired data. Voice Conversion strategies are com- monly restricted to the use of parallel speech corpora. Our propo- sition is based on two main concepts: the modeling of the timbre space based on phonetic information and a simple approximation of the cross-covariance of source-target features. The experimen- tal results based on the mentioned strategy in singing-voice data of the VOCALOID synthesizer showed a conversion performance comparable to that obtained by Maximum-Likelihood, thereby al- lowing us to achieve singer-timbre conversion from real singing performances.

We present in this work a strategy to perform timbre conversion from un- paired source and target data and its application to the singing-voice synthe- sizer VOCALOID to produce sung utterances with a past singer voice. The conversion framework using unpaired data is based on a phoneme-constrained modeling of the timbre space and the assumption of a linear relation between the source and target features. The proposed nonparallel framework resulted in a performance close to the one following the traditional approach based on GMM and paired data. The application to convert an original singer database using sung performances of a past singer observed a successful perception of the past singer’s timbre on the singing-voice utterances performed by VOCALOID.

Abstract—The goal of this project is to convert a given speaker's speech (the Source speaker) into another identified voice (the Target speaker) as well as analysing the face animation of the source to animate a 3D avatar imitating the source facial movements. We assume we have at our disposal a large amount of speech samples from the source and target voices with a reasonable amount of parallel data. Speech and video are processed separately and recombined at the end. Voice conversion is obtained in two steps: a voice ...

With One Voice: Disambiguating Sung and Spoken Voices Through a Composer’s Experience.

Mary Midgley has pointed out how our drive towards specialisation has led us to value the differences between things over their connections. The voice is largely dislodged from educational development and has been divided by isolationist practices in and within music and speech. Technology has allowed microscopic details of the voice to become public property whilst the athletic, full-voiced singer moves ever closer to Lucie Manén’s  “…one timbre, according to their own psychological make-up.” I will discuss my experience of these issues in my work with singers of all styles, actors and the synthetic voice. . In relation to the title of this paper, I propose the disambiguation of spoken and sung words requires re-synthesis.

In this paper we present results of unsupervised cross-lingual speaker adaptation applied to text-to-speech synthesis. The application of our research is the personalisation of speech-to-speech translation in which we employ a HMM statistical framework for both speech recognition and synthesis. This framework provides a logical mechanism to adapt synthesised speech output to the voice of the user by way of speech recognition. In this work we present results of several different unsupervised and cross-lingual adaptation approaches as well as an end-to-end speaker adaptive speech-to-speech translation system. Our experiments show that we can successfully apply speaker adaptation in both unsupervised and cross-lingual scenarios and our proposed algorithms seem to generalise well for several language pairs. We also discuss important future directions including the need for better evaluation metrics.► We examine different approaches for personalising HMM-based speech-to-speech translation systems using speaker adaptation. ► Two adaptation approaches based on ‘pipeline’ and a novel ‘unified’ framework are presented. ► Using both frameworks, unsupervised adaptation of HMM-based TTS is able to preserve speaker similarity in the presence of recognition errors. ► There is a need to better understand perception of speaker similarity and to develop better evaluation metrics for this task.

The NAM-to-speech conversion proposed by Toda and colleagues which converts Non-Audible Murmur (NAM) to audible speech by statistical mapping trained using aligned corpora is a very promising technique, but its performance is still insufficient. In this paper, we present our current work to improve the intelligibility and the naturalness of the synthesized speech converted from whispered speech with this technique. The first system is proposed to improve F0 estimation and voicing decision. A simple neural network is used to detect voiced segments in the whisper while a GMM estimates a continuous melodic contour based on training voiced segments. In the second system, we attempt to integrate visual information for improving both spectral estimation, F0 estimation and voicing decision.

Exploiting a tissue-conductive sensor – a stethoscopic microphone – the system developed at NAIST which converts non-audible murmur (NAM) to audible speech by GMM-based statistical mapping is a very promising technique. The quality of the converted speech is however still insufficient for computer-mediated communication, notably because of the poor estimation of F0 from unvoiced speech and because of impoverished phonetic