Hema A. Murthy

Anand Thyagachandran, Hema A Murthy

Computed Tomography (CT) scans provide a detailed image of the lungs, allowing clinicians to observe the extent of damage caused by COVID-19. The CT severity score (CTSS) based scoring method is used to identify the extent of lung involvement observed on a CT scan. This paper presents a domain knowledge-based pipeline for extracting regions of infection in COVID-19 patients using a combination of image-processing algorithms and a pre-trained UNET model. The severity of the infection is then classified into different categories using an ensemble of three machine-learning models: Extreme Gradient Boosting, Extremely Randomized Trees, and Support Vector Machine. The proposed system was evaluated on a validation dataset in the AI-Enabled Medical Image Analysis Workshop and COVID-19 Diagnosis Competition (AI-MIA-COV19D) and achieved a macro F1 score of 64%. These results demonstrate the potential of combining domain knowledge with machine learning techniques for accurate COVID-19 diagnosis using CT scans. The implementation of the proposed system for severity analysis is available at \textit{https://github.com/aanandt/Enhancing-COVID-19-Severity-Analysis-through-Ensemble-Methods.git }

Anusha Prakash, Arun Kumar, Ashish Seth et al.

Cross-lingual dubbing of lecture videos requires the transcription of the original audio, correction and removal of disfluencies, domain term discovery, text-to-text translation into the target language, chunking of text using target language rhythm, text-to-speech synthesis followed by isochronous lipsyncing to the original video. This task becomes challenging when the source and target languages belong to different language families, resulting in differences in generated audio duration. This is further compounded by the original speaker's rhythm, especially for extempore speech. This paper describes the challenges in regenerating English lecture videos in Indian languages semi-automatically. A prototype is developed for dubbing lectures into 9 Indian languages. A mean-opinion-score (MOS) is obtained for two languages, Hindi and Tamil, on two different courses. The output video is compared with the original video in terms of MOS (1-5) and lip synchronisation with scores of 4.09 and 3.74, respectively. The human effort also reduces by 75%.

Basudha Raje, Sadanand Venkatraman, Nandana TP et al.

This study examines the practical applications and methodological implications of Machine Translation in Indian Languages, specifically Bangla, Malayalam, and Telugu, within emerging translation workflows and in relation to existing evaluation frameworks. The choice of languages prioritized in this study is motivated by a triangulation of linguistic diversity, which illustrates the significance of multilingual accommodation of educational technology under NEP 2020. This is further supported by the largest MOOC portal, i.e., NPTEL, which has served as a corpus to facilitate the arguments presented in this paper. The curation of a spontaneous speech corpora that accounts for lucid delivery of technical concepts, considering the retention of suitable register and lexical choices are crucial in a diverse country like India. The findings of this study highlight metric-specific sensitivity and the challenges of morphologically rich and semantically compact features when tested against surface overlapping metrics.

Gowriprasad R, R Aravind, Hema A Murthy

Tabla is a North Indian percussion instrument used as an accompaniment and an exclusive instrument for solo performances. Tabla solo is intricate and elaborate, exhibiting rhythmic evolution through a sequence of homogeneous sections marked by shared rhythmic characteristics. Each section has a specific structure and name associated with it. Tabla learning and performance in the Indian subcontinent is based on stylistic schools called gharana-s. Several compositions by various composers from different gharana-s are played in each section. This paper addresses the task of segmenting the tabla solo concert into musically meaningful sections. We then assign suitable section labels and recognize gharana-s from the sections. We present a diverse collection of over 38 hours of solo tabla recordings for the task. We motivate the problem and present different challenges and facets of the tasks. Inspired by the distinct musical properties of tabla solo, we compute several rhythmic and timbral features for the segmentation task. This work explores the approach of automatically locating the significant changes in the rhythmic structure by analyzing local self-similarity in an unsupervised manner. We also explore supervised random forest and a convolutional neural network trained on hand-crafted features. Both supervised and unsupervised approaches are also tested on a set of held-out recordings. Segmentation of an audio piece into its structural components and labeling is crucial to many music information retrieval applications like repetitive structure finding, audio summarization, and fast music navigation. This work helps us obtain a comprehensive musical description of the tabla solo concert.

Saish Jaiswal, Shreya Nema, Hema A Murthy et al.

Genomic signal processing has been used successfully in bioinformatics to analyze biomolecular sequences and gain varied insights into DNA structure, gene organization, protein binding, sequence evolution, etc. But challenges remain in finding the appropriate spectral representation of a biomolecular sequence, especially when multiple variable-length sequences need to be handled consistently. In this study, we address this challenge in the context of the well-studied problem of classifying genomic sequences into different taxonomic units (strain, phyla, order, etc.). We propose a novel technique that employs signal processing in tandem with Gaussian mixture models to improve the spectral representation of a sequence and subsequently the taxonomic classification accuracies. The sequences are first transformed into spectra, and projected to a subspace, where sequences belonging to different taxons are better distinguishable. Our method outperforms a similar state-of-the-art method on established benchmark datasets by an absolute margin of 6.06% accuracy.

Utkarsh Pathak, Chandra Sai Krishna Gunda, Anusha Prakash et al.

Text-to-speech (TTS) systems typically require high-quality studio data and accurate transcriptions for training. India has 1369 languages, with 22 official using 13 scripts. Training a TTS system for all these languages, most of which have no digital resources, seems a Herculean task. Our work focuses on zero-shot synthesis, particularly for languages whose scripts and phonotactics come from different families. The novelty of our work is in the augmentation of a shared phone representation and modifying the text parsing rules to match the phonotactics of the target language, thus reducing the synthesiser overhead and enabling rapid adaptation. Intelligible and natural speech was generated for Sanskrit, Maharashtrian and Canara Konkani, Maithili and Kurukh by leveraging linguistic connections across languages with suitable synthesisers. Evaluations confirm the effectiveness of this approach, highlighting its potential to expand speech technology access for under-represented languages.

Mari Ganesh Kumar, Jom Kuriakose, Anand Thyagachandran et al.

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.

Nauman Dawalatabad, Jilt Sebastian, Jom Kuriakose et al.

Instrument separation in an ensemble is a challenging task. In this work, we address the problem of separating the percussive voices in the taniavartanam segments of Carnatic music. In taniavartanam, a number of percussive instruments play together or in tandem. Separation of instruments in regions where only one percussion is present leads to interference and artifacts at the output, as source separation algorithms assume the presence of multiple percussive voices throughout the audio segment. We prevent this by first subjecting the taniavartanam to diarization. This process results in homogeneous clusters consisting of segments of either a single voice or multiple voices. A cluster of segments with multiple voices is identified using the Gaussian mixture model (GMM), which is then subjected to source separation. A deep recurrent neural network (DRNN) based approach is used to separate the multiple instrument segments. The effectiveness of the proposed system is evaluated on a standard Carnatic music dataset. The proposed approach provides close-to-oracle performance for non-overlapping segments and a significant improvement over traditional separation schemes.

Vinay Kumar Verma, Ashish Mishra, Anubha Pandey et al.

We present a meta-learning based generative model for zero-shot learning (ZSL) towards a challenging setting when the number of training examples from each \emph{seen} class is very few. This setup contrasts with the conventional ZSL approaches, where training typically assumes the availability of a sufficiently large number of training examples from each of the seen classes. The proposed approach leverages meta-learning to train a deep generative model that integrates variational autoencoder and generative adversarial networks. We propose a novel task distribution where meta-train and meta-validation classes are disjoint to simulate the ZSL behaviour in training. Once trained, the model can generate synthetic examples from seen and unseen classes. Synthesize samples can then be used to train the ZSL framework in a supervised manner. The meta-learner enables our model to generates high-fidelity samples using only a small number of training examples from seen classes. We conduct extensive experiments and ablation studies on four benchmark datasets of ZSL and observe that the proposed model outperforms state-of-the-art approaches by a significant margin when the number of examples per seen class is very small.

Rini A Sharon, Hema A Murthy

Translation of imagined speech electroencephalogram(EEG) into human understandable commands greatly facilitates the design of naturalistic brain computer interfaces. To achieve improved imagined speech unit classification, this work aims to profit from the parallel information contained in multi-phasal EEG data recorded while speaking, imagining and performing articulatory movements corresponding to specific speech units. A bi-phase common representation learning module using neural networks is designed to model the correlation and reproducibility between an analysis phase and a support phase. The trained Correlation Network is then employed to extract discriminative features of the analysis phase. These features are further classified into five binary phonological categories using machine learning models such as Gaussian mixture based hidden Markov model and deep neural networks. The proposed approach further handles the non-availability of multi-phasal data during decoding. Topographic visualizations along with result-based inferences suggest that the multi-phasal correlation modelling approach proposed in the paper enhances imagined-speech EEG recognition performance.

Rini A Sharon, Hema A Murthy

Speech cognition bears potential application as a brain computer interface that can improve the quality of life for the otherwise communication impaired people. While speech and resting state EEG are popularly studied, here we attempt to explore a "non-speech"(NS) state of brain activity corresponding to the silence regions of speech audio. Firstly, speech perception is studied to inspect the existence of such a state, followed by its identification in speech imagination. Analogous to how voice activity detection is employed to enhance the performance of speech recognition, the EEG state activity detection protocol implemented here is applied to boost the confidence of imagined speech EEG decoding. Classification of speech and NS state is done using two datasets collected from laboratory-based and commercial-based devices. The state sequential information thus obtained is further utilized to reduce the search space of imagined EEG unit recognition. Temporal signal structures and topographic maps of NS states are visualized across subjects and sessions. The recognition performance and the visual distinction observed demonstrates the existence of silence signatures in EEG.

Karthik Pandia D S, Anusha Prakash, Mano Ranjith Kumar et al.

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.

Karthik Pandia D S, Hema A Murthy

Zerospeech synthesis is the task of building vocabulary independent speech synthesis systems, where transcriptions are not available for training data. It is, therefore, necessary to convert training data into a sequence of fundamental acoustic units that can be used for synthesis during the test. This paper attempts to discover, and model perceptual acoustic units consisting of steady-state, and transient regions in speech. The transients roughly correspond to CV, VC units, while the steady-state corresponds to sonorants and fricatives. The speech signal is first preprocessed by segmenting the same into CVC-like units using a short-term energy-like contour. These CVC segments are clustered using a connected components-based graph clustering technique. The clustered CVC segments are initialized such that the onset (CV) and decays (VC) correspond to transients, and the rhyme corresponds to steady-states. Following this initialization, the units are allowed to re-organise on the continuous speech into a final set of AUs in an HMM-GMM framework. AU sequences thus obtained are used to train synthesis models. The performance of the proposed approach is evaluated on the Zerospeech 2019 challenge database. Subjective and objective scores show that reasonably good quality synthesis with low bit rate encoding can be achieved using the proposed AUs.

Anubha Pandey, Ashish Mishra, Vinay Kumar Verma et al.

Conventional approaches to Sketch-Based Image Retrieval (SBIR) assume that the data of all the classes are available during training. The assumption may not always be practical since the data of a few classes may be unavailable, or the classes may not appear at the time of training. Zero-Shot Sketch-Based Image Retrieval (ZS-SBIR) relaxes this constraint and allows the algorithm to handle previously unseen classes during the test. This paper proposes a generative approach based on the Stacked Adversarial Network (SAN) and the advantage of Siamese Network (SN) for ZS-SBIR. While SAN generates a high-quality sample, SN learns a better distance metric compared to that of the nearest neighbor search. The capability of the generative model to synthesize image features based on the sketch reduces the SBIR problem to that of an image-to-image retrieval problem. We evaluate the efficacy of our proposed approach on TU-Berlin, and Sketchy database in both standard ZSL and generalized ZSL setting. The proposed method yields a significant improvement in standard ZSL as well as in a more challenging generalized ZSL setting (GZSL) for SBIR.

Nauman Dawalatabad, Srikanth Madikeri, C Chandra Sekhar et al.

The two-pass information bottleneck (TPIB) based speaker diarization system operates independently on different conversational recordings. TPIB system does not consider previously learned speaker discriminative information while diarizing new conversations. Hence, the real time factor (RTF) of TPIB system is high owing to the training time required for the artificial neural network (ANN). This paper attempts to improve the RTF of the TPIB system using an incremental transfer learning approach where the parameters learned by the ANN from other conversations are updated using current conversation rather than learning parameters from scratch. This reduces the RTF significantly. The effectiveness of the proposed approach compared to the baseline IB and the TPIB systems is demonstrated on standard NIST and AMI conversational meeting datasets. With a minor degradation in performance, the proposed system shows a significant improvement of 33.07% and 24.45% in RTF with respect to TPIB system on the NIST RT-04Eval and AMI-1 datasets, respectively.

Venkata Subramanian Viraraghavan, Arpan Pal, R Aravind et al.

Gamakas are an integral aspect of Carnatic Music, a form of classical music prevalent in South India. They are used in ragas, which may be seen as melodic scales and/or a set of characteristic melodic phrases. Gamakas exhibit continuous pitch variation often spanning several semitones. In this paper, we study how gamakas scale with tempo and propose a novel approach to change the tempo of Carnatic music pieces. The music signal is viewed as consisting of constant-pitch segments and transients. The transients show continuous pitch variation and we consider their analyses from a theoretical stand-point. We next observe the non-uniform ratios of time-scaling of constant-pitch segments, transients and silence in excerpts from nine concert renditions of varnams in six ragas. The results indicate that the changing tempo of Carnatic music does not change the duration of transients significantly. We report listening tests on our algorithm to slow down Carnatic music that is consistent with this observation.

Ashish Mishra, M Shiva Krishna Reddy, Anurag Mittal et al.

Zero shot learning in Image Classification refers to the setting where images from some novel classes are absent in the training data but other information such as natural language descriptions or attribute vectors of the classes are available. This setting is important in the real world since one may not be able to obtain images of all the possible classes at training. While previous approaches have tried to model the relationship between the class attribute space and the image space via some kind of a transfer function in order to model the image space correspondingly to an unseen class, we take a different approach and try to generate the samples from the given attributes, using a conditional variational autoencoder, and use the generated samples for classification of the unseen classes. By extensive testing on four benchmark datasets, we show that our model outperforms the state of the art, particularly in the more realistic generalized setting, where the training classes can also appear at the test time along with the novel classes.

Rajeev Rajan, Hema A. Murthy

Phase processing has been replaced by group delay processing for the extraction of source and system parameters from speech. Group delay functions are ill-behaved when the transfer function has zeros that are close to unit circle in the z-domain. The modified group delay function addresses this problem and has been successfully used for formant and monopitch estimation. In this paper, modified group delay functions are used for multipitch estimation in concurrent speech. The power spectrum of the speech is first flattened in order to annihilate the system characteristics, while retaining the source characteristics. Group delay analysis on this flattened spectrum picks the predominant pitch in the first pass and a comb filter is used to filter out the estimated pitch along with its harmonics. The residual spectrum is again analyzed for the next candidate pitch estimate in the second pass. The final pitch trajectories of the constituent speech utterances are formed using pitch grouping and post processing techniques. The performance of the proposed algorithm was evaluated on standard datasets using two metrics; pitch accuracy and standard deviation of fine pitch error. Our results show that the proposed algorithm is a promising pitch detection method in multipitch environment for real speech recordings.