End-to-end (E2E) systems synthesise high-quality speech, but this typically requires a large amount of data. As E2E synthesis progressed from Tacotron to FastSpeech2, it became evident that features representing prosody, particularly subword durations, are important for error-free synthesis. Variants of FastSpeech use a teacher model or forced alignments for training. This paper uses signal processing cues in tandem with forced alignment to produce accurate phone boundaries for the training data. As a result of better duration modelling, good-quality synthesisers are developed. Evaluations indicate that systems developed using the proposed signal processing-aided approach are better than systems developed using other alignment approaches, especially in low-resource scenarios. Our systems also outperform the existing best TTS systems available for 13 Indian languages.

Building end-to-end speech synthesisers for Indian languages is challenging, given the lack of adequate clean training data and multiple grapheme representations across languages. This work explores the importance of training multilingual and multi-speaker text-to-speech (TTS) systems based onlanguage families. The objective is to exploit the phonotactic properties of language families, where small amounts of accurately transcribed data across languages can be pooled together to train TTS systems. These systems can then be adapted to new languages belonging to the same family in extremely low-resource scenarios. TTS systems are trained separately for Indo-Aryan and Dravidian language families, and their performance is compared to that of a combined Indo-Aryan+Dravidian voice. We also investigate the amount of training data required for a language in a multilingual setting. Same-family and cross-family synthesis and adaptation to unseen languages are analysed. The analyses show that language family-wise training of Indic systems is the way forward for the Indian subcontinent, where a large number of languages are spoken.

Conversational speech has always been challenging in the context of text-to-speech synthesis (TTS). Most speech synthesis systems are trained on read speech data recorded in a studio environment. But, the intelligibility of TTS systems degrades drastically when using conversational speech. The proposed work attempts to perform extensive analysis on the issues in dealing with conversational speech compared to read speech. As an application, we try to dub the lectures available in English into an Indian language (Hindi) in the original speaker's voice. The task is difficult as classroom lectures are extempore, with variations in speaking rate, and contain speaker mannerisms that lead to disfluencies. We analyze the capability of end-to-end TTS systems in modeling lecture-based data. Based on the analysis, an attempt is made to adapt “read speech TTS system” using conversational speech data to produce lectures in the original speaker's voice.

India is home to multiple languages, and training automatic speech recognition (ASR) systems for languages is challenging. Over time, each language has adopted words from other languages, such as English, leading to code-mixing. Most Indian languages also have their own unique scripts, which poses a major limitation in training multilingual and code-switching ASR systems. Inspired by results in text-to-speech synthesis, in this work, we use an in-house rule-based phoneme-level common label set (CLS) representation to train multilingual and code-switching ASR for Indian languages. We propose two end-to-end (E2E) ASR systems. In the first system, the E2E model is trained on the CLS representation, and we use a novel data-driven back-end to recover the native language script. In the second system, we propose a modification to the E2E model, wherein the CLS representation and the native language characters are used simultaneously for training. We show our results on the multilingual and code-switching tasks of the Indic ASR Challenge 2021. Our best results achieve 6% and 5% improvement (approx) in word error rate over the baseline system for the multilingual and code-switching tasks, respectively, on the challenge development data.

Building text-to-speech (TTS) synthesisers for Indian languages is a difficult task owing to a large number of active languages. Indian languages can be classified into a finite set of families, prominent among them, Indo-Aryan and Dravidian. The proposed work exploits this property to build a generic TTS system using multiple languages from the same family in an end-to-end framework. Generic systems are quite robust as they are capable of capturing a variety of phonotactics across languages. These systems are then adapted to a new language in the same family using small amounts of adaptation data. Experiments indicate that good quality TTS systems can be built using only 7 minutes of adaptation data. An average degradation mean opinion score of 3.98 is obtained for the adapted TTSes. Extensive analysis of systematic interactions between languages in the generic TTSes is carried out. x-vectors are included as speaker embedding to synthesise text in a particular speaker's voice. An interesting observation is that the prosody of the target speaker's voice is preserved. These results are quite promising as they indicate the capability of generic TTSes to handle speaker and language switching seamlessly, along with the ease of adaptation to a new language.

A Spoken dialogue system for an unseen language is referred to as Zero resource speech. It is especially beneficial for developing applications for languages that have low digital resources. Zero resource speech synthesis is the task of building text-to-speech (TTS) models in the absence of transcriptions. In this work, speech is modelled as a sequence of transient and steady-state acoustic units, and a unique set of acoustic units is discovered by iterative training. Using the acoustic unit sequence, TTS models are trained. The main goal of this work is to improve the synthesis quality of zero resource TTS system. Four different systems are proposed. All the systems consist of three stages: unit discovery, followed by unit sequence to spectrogram mapping, and finally spectrogram to speech inversion. Modifications are proposed to the spectrogram mapping stage. These modifications include training the mapping on voice data, using x-vectors to improve the mapping, two-stage learning, and gender-specific modelling. Evaluation of the proposed systems in the Zerospeech 2020 challenge shows that quite good quality synthesis can be achieved.