Mathew Magimai. -Doss

Sargam Vyas, Bogdan Vlasenko, André Mayoraz et al.

With advancements in multimodal communication technologies, remote learning environments such as, distance universities are increasing. Remote learning typically happens asynchronously. As a consequence, unlike face-to-face in-person classroom teaching, this lacks availability of sufficient emotional cues for making learning a pleasant experience. Motivated by advances made in the paralinguistic speech processing community on emotion prediction, in this paper we explore use of speech for sensing students' emotions by building upon speech-based self-control tasks developed to aid effective remote learning. More precisely, we investigate: (a) whether speech acquired through self-control tasks exhibit perceptible variation along valence, arousal, and dominance dimensions? and (b) whether those dimensional emotion variations can be automatically predicted? We address these two research questions by developing a dataset containing spontaneous monologue speech acquired as open responses to self-control tasks and by carrying out subjective listener evaluations and automatic dimensional emotion prediction studies on that dataset. Our investigations indicate that speech-based self-control tasks can be a means to sense student emotion in remote learning environment. This opens potential venues to seamlessly integrate paralinguistic speech processing technologies in the remote learning loop for enhancing learning experiences through instructional design and feedback generation.

Maryam Naderi, Enno Hermann, Alexandre Nanchen et al.

As large language models (LLMs) grow in parameter size and capabilities, such as interaction through prompting, they open up new ways of interfacing with automatic speech recognition (ASR) systems beyond rescoring n-best lists. This work investigates post-hoc correction of ASR transcripts with LLMs. To avoid introducing errors into likely accurate transcripts, we propose a range of confidence-based filtering methods. Our results indicate that this can improve the performance of less competitive ASR systems.

Karl El Hajal, Ajinkya Kulkarni, Enno Hermann et al.

While recent zero-shot multi-speaker text-to-speech (TTS) models achieve impressive results, they typically rely on extensive transcribed speech datasets from numerous speakers and intricate training pipelines. Meanwhile, self-supervised learning (SSL) speech features have emerged as effective intermediate representations for TTS. Further, SSL features from different speakers that are linearly close share phonetic information while maintaining individual speaker identity. In this study, we introduce kNN-TTS, a simple and effective framework for zero-shot multi-speaker TTS using retrieval methods which leverage the linear relationships between SSL features. Objective and subjective evaluations show that our models, trained on transcribed speech from a single speaker only, achieve performance comparable to state-of-the-art models that are trained on significantly larger training datasets. The low training data requirements mean that kNN-TTS is well suited for the development of multi-speaker TTS systems for low-resource domains and languages. We also introduce an interpolation parameter which enables fine-grained voice morphing. Demo samples are available at https://idiap.github.io/knn-tts

Eklavya Sarkar, Mathew Magimai. -Doss

Marmoset monkeys encode vital information in their calls and serve as a surrogate model for neuro-biologists to understand the evolutionary origins of human vocal communication. Traditionally analyzed with signal processing-based features, recent approaches have utilized self-supervised models pre-trained on human speech for feature extraction, capitalizing on their ability to learn a signal's intrinsic structure independently of its acoustic domain. However, the utility of such foundation models remains unclear for marmoset call analysis in terms of multi-class classification, bandwidth, and pre-training domain. This study assesses feature representations derived from speech and general audio domains, across pre-training bandwidths of 4, 8, and 16 kHz for marmoset call-type and caller classification tasks. Results show that models with higher bandwidth improve performance, and pre-training on speech or general audio yields comparable results, improving over a spectral baseline.

Imen Ben Mahmoud, Eklavya Sarkar, Marta Manser et al.

Understanding evolution of vocal communication in social animals is an important research problem. In that context, beyond humans, there is an interest in analyzing vocalizations of other social animals such as, meerkats, marmosets, apes. While existing approaches address vocalizations of certain species, a reliable method tailored for meerkat calls is lacking. To that extent, this paper investigates feature representations for automatic meerkat vocalization analysis. Both traditional signal processing-based representations and data-driven representations facilitated by advances in deep learning are explored. Call type classification studies conducted on two data sets reveal that feature extraction methods developed for human speech processing can be effectively employed for automatic meerkat call analysis.

Eklavya Sarkar, Mathew Magimai. -Doss

Animal vocalizations contain sequential structures that carry important communicative information, yet most computational bioacoustics studies average the extracted frame-level features across the temporal axis, discarding the order of the sub-units within a vocalization. This paper investigates whether discrete acoustic token sequences, derived through vector quantization and gumbel-softmax vector quantization of extracted self-supervised speech model representations can effectively capture and leverage temporal information. To that end, pairwise distance analysis of token sequences generated from HuBERT embeddings shows that they can discriminate call-types and callers across four bioacoustics datasets. Sequence classification experiments using $k$-Nearest Neighbour with Levenshtein distance show that the vector-quantized token sequences yield reasonable call-type and caller classification performances, and hold promise as alternative feature representations towards leveraging sequential information in animal vocalizations.

Karl El Hajal, Enno Hermann, Sevada Hovsepyan et al.

Automatic speech recognition (ASR) systems struggle with dysarthric speech due to high inter-speaker variability and slow speaking rates. To address this, we explore dysarthric-to-healthy speech conversion for improved ASR performance. Our approach extends the Rhythm and Voice (RnV) conversion framework by introducing a syllable-based rhythm modeling method suited for dysarthric speech. We assess its impact on ASR by training LF-MMI models and fine-tuning Whisper on converted speech. Experiments on the Torgo corpus reveal that LF-MMI achieves significant word error rate reductions, especially for more severe cases of dysarthria, while fine-tuning Whisper on converted data has minimal effect on its performance. These results highlight the potential of unsupervised rhythm and voice conversion for dysarthric ASR. Code available at: https://github.com/idiap/RnV

Panagiotis Theocharopoulos, Ajinkya Kulkarni, Mathew Magimai. -Doss

Large language models (LLMs) are increasingly considered for use in high-impact workflows, including academic peer review. However, LLMs are vulnerable to document-level hidden prompt injection attacks. In this work, we construct a dataset of approximately 500 real academic papers accepted to ICML and evaluate the effect of embedding hidden adversarial prompts within these documents. Each paper is injected with semantically equivalent instructions in four different languages and reviewed using an LLM. We find that prompt injection induces substantial changes in review scores and accept/reject decisions for English, Japanese, and Chinese injections, while Arabic injections produce little to no effect. These results highlight the susceptibility of LLM-based reviewing systems to document-level prompt injection and reveal notable differences in vulnerability across languages.

Eklavya Sarkar, Mathew Magimai. -Doss

Self-supervised learning (SSL) foundation models have emerged as powerful, domain-agnostic, general-purpose feature extractors applicable to a wide range of tasks. Such models pre-trained on human speech have demonstrated high transferability for bioacoustic processing. This paper investigates (i) whether SSL models pre-trained directly on animal vocalizations offer a significant advantage over those pre-trained on speech, and (ii) whether fine-tuning speech-pretrained models on automatic speech recognition (ASR) tasks can enhance bioacoustic classification. We conduct a comparative analysis using three diverse bioacoustic datasets and two different bioacoustic tasks. Results indicate that pre-training on bioacoustic data provides only marginal improvements over speech-pretrained models, with comparable performance in most scenarios. Fine-tuning on ASR tasks yields mixed outcomes, suggesting that the general-purpose representations learned during SSL pre-training are already well-suited for bioacoustic tasks. These findings highlight the robustness of speech-pretrained SSL models for bioacoustics and imply that extensive fine-tuning may not be necessary for optimal performance.

Karl El Hajal, Enno Hermann, Ajinkya Kulkarni et al.

Automatic speech recognition (ASR) systems are well known to perform poorly on dysarthric speech. Previous works have addressed this by speaking rate modification to reduce the mismatch with typical speech. Unfortunately, these approaches rely on transcribed speech data to estimate speaking rates and phoneme durations, which might not be available for unseen speakers. Therefore, we combine unsupervised rhythm and voice conversion methods based on self-supervised speech representations to map dysarthric to typical speech. We evaluate the outputs with a large ASR model pre-trained on healthy speech without further fine-tuning and find that the proposed rhythm conversion especially improves performance for speakers of the Torgo corpus with more severe cases of dysarthria. Code and audio samples are available at https://idiap.github.io/RnV .

Ajinkya Kulkarni, Sandipana Dowerah, Tanel Alumae et al.

Audio deepfakes are acquiring an unprecedented level of realism with advanced AI. While current research focuses on discerning real speech from spoofed speech, tracing the source system is equally crucial. This work proposes a novel audio source tracing system combining deep metric multi-class N-pair loss with Real Emphasis and Fake Dispersion framework, a Conformer classification network, and ensemble score-embedding fusion. The N-pair loss improves discriminative ability, while Real Emphasis and Fake Dispersion enhance robustness by focusing on differentiating real and fake speech patterns. The Conformer network captures both global and local dependencies in the audio signal, crucial for source tracing. The proposed ensemble score-embedding fusion shows an optimal trade-off between in-domain and out-of-domain source tracing scenarios. We evaluate our method using Frechet Distance and standard metrics, demonstrating superior performance in source tracing over the baseline system.

Gasser Elbanna, Zohreh Mostaani, Mathew Magimai. -Doss

Accurately predicting heart activity and other biological signals is crucial for diagnosis and monitoring. Given that speech is an outcome of multiple physiological systems, a significant body of work studied the acoustic correlates of heart activity. Recently, self-supervised models have excelled in speech-related tasks compared to traditional acoustic methods. However, the robustness of data-driven representations in predicting heart activity remained unexplored. In this study, we demonstrate that self-supervised speech models outperform acoustic features in predicting heart activity parameters. We also emphasize the impact of individual variability on model generalizability. These findings underscore the value of data-driven representations in such tasks and the need for more speech-based physiological data to mitigate speaker-related challenges.

Eklavya Sarkar, Mathew Magimai. -Doss

Self-supervised learning (SSL) models use only the intrinsic structure of a given signal, independent of its acoustic domain, to extract essential information from the input to an embedding space. This implies that the utility of such representations is not limited to modeling human speech alone. Building on this understanding, this paper explores the cross-transferability of SSL neural representations learned from human speech to analyze bio-acoustic signals. We conduct a caller discrimination analysis and a caller detection study on Marmoset vocalizations using eleven SSL models pre-trained with various pretext tasks. The results show that the embedding spaces carry meaningful caller information and can successfully distinguish the individual identities of Marmoset callers without fine-tuning. This demonstrates that representations pre-trained on human speech can be effectively applied to the bio-acoustics domain, providing valuable insights for future investigations in this field.

Bence Mark Halpern, Julian Fritsch, Enno Hermann et al.

The development of pathological speech systems is currently hindered by the lack of a standardised objective evaluation framework. In this work, (1) we utilise existing detection and analysis techniques to propose a general framework for the consistent evaluation of synthetic pathological speech. This framework evaluates the voice quality and the intelligibility aspects of speech and is shown to be complementary using our experiments. (2) Using our proposed evaluation framework, we develop and test a dysarthric voice conversion system (VC) using CycleGAN-VC and a PSOLA-based speech rate modification technique. We show that the developed system is able to synthesise dysarthric speech with different levels of speech intelligibility.

Dimitri Palaz, Ronan Collobert, Mathew Magimai. -Doss

Most phoneme recognition state-of-the-art systems rely on a classical neural network classifiers, fed with highly tuned features, such as MFCC or PLP features. Recent advances in ``deep learning'' approaches questioned such systems, but while some attempts were made with simpler features such as spectrograms, state-of-the-art systems still rely on MFCCs. This might be viewed as a kind of failure from deep learning approaches, which are often claimed to have the ability to train with raw signals, alleviating the need of hand-crafted features. In this paper, we investigate a convolutional neural network approach for raw speech signals. While convolutional architectures got tremendous success in computer vision or text processing, they seem to have been let down in the past recent years in the speech processing field. We show that it is possible to learn an end-to-end phoneme sequence classifier system directly from raw signal, with similar performance on the TIMIT and WSJ datasets than existing systems based on MFCC, questioning the need of complex hand-crafted features on large datasets.

Dimitri Palaz, Ronan Collobert, Mathew Magimai. -Doss

In hybrid hidden Markov model/artificial neural networks (HMM/ANN) automatic speech recognition (ASR) system, the phoneme class conditional probabilities are estimated by first extracting acoustic features from the speech signal based on prior knowledge such as, speech perception or/and speech production knowledge, and, then modeling the acoustic features with an ANN. Recent advances in machine learning techniques, more specifically in the field of image processing and text processing, have shown that such divide and conquer strategy (i.e., separating feature extraction and modeling steps) may not be necessary. Motivated from these studies, in the framework of convolutional neural networks (CNNs), this paper investigates a novel approach, where the input to the ANN is raw speech signal and the output is phoneme class conditional probability estimates. On TIMIT phoneme recognition task, we study different ANN architectures to show the benefit of CNNs and compare the proposed approach against conventional approach where, spectral-based feature MFCC is extracted and modeled by a multilayer perceptron. Our studies show that the proposed approach can yield comparable or better phoneme recognition performance when compared to the conventional approach. It indicates that CNNs can learn features relevant for phoneme classification automatically from the raw speech signal.