Dipjyoti Paul

Dipjyoti Paul, Muhammed PV Shifas, Yannis Pantazis et al.

The increased adoption of digital assistants makes text-to-speech (TTS) synthesis systems an indispensable feature of modern mobile devices. It is hence desirable to build a system capable of generating highly intelligible speech in the presence of noise. Past studies have investigated style conversion in TTS synthesis, yet degraded synthesized quality often leads to worse intelligibility. To overcome such limitations, we proposed a novel transfer learning approach using Tacotron and WaveRNN based TTS synthesis. The proposed speech system exploits two modification strategies: (a) Lombard speaking style data and (b) Spectral Shaping and Dynamic Range Compression (SSDRC) which has been shown to provide high intelligibility gains by redistributing the signal energy on the time-frequency domain. We refer to this extension as Lombard-SSDRC TTS system. Intelligibility enhancement as quantified by the Intelligibility in Bits (SIIB-Gauss) measure shows that the proposed Lombard-SSDRC TTS system shows significant relative improvement between 110% and 130% in speech-shaped noise (SSN), and 47% to 140% in competing-speaker noise (CSN) against the state-of-the-art TTS approach. Additional subjective evaluation shows that Lombard-SSDRC TTS successfully increases the speech intelligibility with relative improvement of 455% for SSN and 104% for CSN in median keyword correction rate compared to the baseline TTS method.

Dipjyoti Paul, Yannis Pantazis, Yannis Stylianou

Recent advancements in deep learning led to human-level performance in single-speaker speech synthesis. However, there are still limitations in terms of speech quality when generalizing those systems into multiple-speaker models especially for unseen speakers and unseen recording qualities. For instance, conventional neural vocoders are adjusted to the training speaker and have poor generalization capabilities to unseen speakers. In this work, we propose a variant of WaveRNN, referred to as speaker conditional WaveRNN (SC-WaveRNN). We target towards the development of an efficient universal vocoder even for unseen speakers and recording conditions. In contrast to standard WaveRNN, SC-WaveRNN exploits additional information given in the form of speaker embeddings. Using publicly-available data for training, SC-WaveRNN achieves significantly better performance over baseline WaveRNN on both subjective and objective metrics. In MOS, SC-WaveRNN achieves an improvement of about 23% for seen speaker and seen recording condition and up to 95% for unseen speaker and unseen condition. Finally, we extend our work by implementing a multi-speaker text-to-speech (TTS) synthesis similar to zero-shot speaker adaptation. In terms of performance, our system has been preferred over the baseline TTS system by 60% over 15.5% and by 60.9% over 32.6%, for seen and unseen speakers, respectively.

Yannis Pantazis, Dipjyoti Paul, Michail Fasoulakis et al.

In this paper, we propose a novel loss function for training Generative Adversarial Networks (GANs) aiming towards deeper theoretical understanding as well as improved stability and performance for the underlying optimization problem. The new loss function is based on cumulant generating functions giving rise to \emph{Cumulant GAN}. Relying on a recently-derived variational formula, we show that the corresponding optimization problem is equivalent to R{é}nyi divergence minimization, thus offering a (partially) unified perspective of GAN losses: the R{é}nyi family encompasses Kullback-Leibler divergence (KLD), reverse KLD, Hellinger distance and $χ^2$-divergence. Wasserstein GAN is also a member of cumulant GAN. In terms of stability, we rigorously prove the linear convergence of cumulant GAN to the Nash equilibrium for a linear discriminator, Gaussian distributions and the standard gradient descent ascent algorithm. Finally, we experimentally demonstrate that image generation is more robust relative to Wasserstein GAN and it is substantially improved in terms of both inception score and Fréchet inception distance when both weaker and stronger discriminators are considered.

Dipjyoti Paul, Md Sahidullah, Goutam Saha

Voice-based biometric systems are highly prone to spoofing attacks. Recently, various countermeasures have been developed for detecting different kinds of attacks such as replay, speech synthesis (SS) and voice conversion (VC). Most of the existing studies are conducted with a specific training set defined by the evaluation protocol. However, for realistic scenarios, selecting appropriate training data is an open challenge for the system administrator. Motivated by this practical concern, this work investigates the generalization capability of spoofing countermeasures in restricted training conditions where speech from a broad attack types are left out in the training database. We demonstrate that different spoofing types have considerably different generalization capabilities. For this study, we analyze the performance using two kinds of features, mel-frequency cepstral coefficients (MFCCs) which are considered as baseline and recently proposed constant Q cepstral coefficients (CQCCs). The experiments are conducted with standard Gaussian mixture model - maximum likelihood (GMM-ML) classifier on two recently released spoofing corpora: ASVspoof 2015 and BTAS 2016 that includes cross-corpora performance analysis. Feature-level analysis suggests that static and dynamic coefficients of spectral features, both are important for detecting spoofing attacks in the real-life condition.

Yannis Pantazis, Dipjyoti Paul, Michail Fasoulakis et al.

The impressive success of Generative Adversarial Networks (GANs) is often overshadowed by the difficulties in their training. Despite the continuous efforts and improvements, there are still open issues regarding their convergence properties. In this paper, we propose a simple training variation where suitable weights are defined and assist the training of the Generator. We provide theoretical arguments why the proposed algorithm is better than the baseline training in the sense of speeding up the training process and of creating a stronger Generator. Performance results showed that the new algorithm is more accurate in both synthetic and image datasets resulting in improvements ranging between 5% and 50%.

Monisankha Pal, Dipjyoti Paul, Md Sahidullah et al.

Most of the existing studies on voice conversion (VC) are conducted in acoustically matched conditions between source and target signal. However, the robustness of VC methods in presence of mismatch remains unknown. In this paper, we report a comparative analysis of different VC techniques under mismatched conditions. The extensive experiments with five different VC techniques on CMU ARCTIC corpus suggest that performance of VC methods substantially degrades in noisy conditions. We have found that bilinear frequency warping with amplitude scaling (BLFWAS) outperforms other methods in most of the noisy conditions. We further explore the suitability of different speech enhancement techniques for robust conversion. The objective evaluation results indicate that spectral subtraction and log minimum mean square error (logMMSE) based speech enhancement techniques can be used to improve the performance in specific noisy conditions.

Dipjyoti Paul, Monisankha Pal, Goutam Saha

Now-a-days, speech-based biometric systems such as automatic speaker verification (ASV) are highly prone to spoofing attacks by an imposture. With recent development in various voice conversion (VC) and speech synthesis (SS) algorithms, these spoofing attacks can pose a serious potential threat to the current state-of-the-art ASV systems. To impede such attacks and enhance the security of the ASV systems, the development of efficient anti-spoofing algorithms is essential that can differentiate synthetic or converted speech from natural or human speech. In this paper, we propose a set of novel speech features for detecting spoofing attacks. The proposed features are computed using alternative frequency-warping technique and formant-specific block transformation of filter bank log energies. We have evaluated existing and proposed features against several kinds of synthetic speech data from ASVspoof 2015 corpora. The results show that the proposed techniques outperform existing approaches for various spoofing attack detection task. The techniques investigated in this paper can also accurately classify natural and synthetic speech as equal error rates (EERs) of 0% have been achieved.