Self-supervised learned models have been found to be very effective for certain speech tasks such as automatic speech recognition, speaker identification, keyword spotting and others. While the features are undeniably useful in speech recognition and associated tasks, their utility in speech enhancement systems is yet to be firmly established, and perhaps not properly understood. In this paper, we investigate the uses of SSL representations for single-channel speech enhancement in challenging conditions and find that they add very little value for the enhancement task. Our constraints are designed around on-device real-time speech enhancement -- model is causal, the compute footprint is small. Additionally, we focus on low SNR conditions where such models struggle to provide good enhancement. In order to systematically examine how SSL representations impact performance of such enhancement models, we propose a variety of techniques to utilize these embeddings which include different forms of knowledge-distillation and pre-training.
In the wake of the surging tide of deep learning over the past decade, Automatic Speech Recognition (ASR) has garnered substantial attention, leading to the emergence of numerous publicly accessible ASR systems that are actively being integrated into our daily lives. Nonetheless, the impartial and replicable evaluation of these ASR systems encounters challenges due to various crucial subtleties. In this paper we introduce the SpeechColab Leaderboard, a general-purpose, open-source platform designed for ASR evaluation. With this platform: (i) We report a comprehensive benchmark, unveiling the current state-of-the-art panorama for ASR systems, covering both open-source models and industrial commercial services. (ii) We quantize how distinct nuances in the scoring pipeline influence the final benchmark outcomes. These include nuances related to capitalization, punctuation, interjection, contraction, synonym usage, compound words, etc. These issues have gained prominence in the context of the transition towards an End-to-End future. (iii) We propose a practical modification to the conventional Token-Error-Rate (TER) evaluation metric, with inspirations from Kolmogorov complexity and Normalized Information Distance (NID). This adaptation, called modified-TER (mTER), achieves proper normalization and symmetrical treatment of reference and hypothesis. By leveraging this platform as a large-scale testing ground, this study demonstrates the robustness and backward compatibility of mTER when compared to TER. The SpeechColab Leaderboard is accessible at https://github.com/SpeechColab/Leaderboard
Decoding language from brain dynamics is an important open direction in the realm of brain-computer interface (BCI), especially considering the rapid growth of large language models. Compared to invasive-based signals which require electrode implantation surgery, non-invasive neural signals (e.g. EEG, MEG) have attracted increasing attention considering their safety and generality. However, the exploration is not adequate in three aspects: 1) previous methods mainly focus on EEG but none of the previous works address this problem on MEG with better signal quality; 2) prior works have predominantly used ``teacher-forcing" during generative decoding, which is impractical; 3) prior works are mostly ``BART-based" not fully auto-regressive, which performs better in other sequence tasks. In this paper, we explore the brain-to-text translation of MEG signals in a speech-decoding formation. Here we are the first to investigate a cross-attention-based ``whisper" model for generating text directly from MEG signals without teacher forcing. Our model achieves impressive BLEU-1 scores of 60.30 and 52.89 without pretraining \& teacher-forcing on two major datasets (\textit{GWilliams} and \textit{Schoffelen}). This paper conducts a comprehensive review to understand how speech decoding formation performs on the neural decoding tasks, including pretraining initialization, training \& evaluation set splitting, augmentation, and scaling law.
Collecting high-quality studio recordings of audio is challenging, which limits the language coverage of text-to-speech (TTS) systems. This paper proposes a framework for scaling a multilingual TTS model to 100+ languages using found data without supervision. The proposed framework combines speech-text encoder pretraining with unsupervised training using untranscribed speech and unspoken text data sources, thereby leveraging massively multilingual joint speech and text representation learning. Without any transcribed speech in a new language, this TTS model can generate intelligible speech in >30 unseen languages (CER difference of <10% to ground truth). With just 15 minutes of transcribed, found data, we can reduce the intelligibility difference to 1% or less from the ground-truth, and achieve naturalness scores that match the ground-truth in several languages.
Recently, and under the umbrella of Responsible AI, efforts have been made to develop gender-ambiguous synthetic speech to represent with a single voice all individuals in the gender spectrum. However, research efforts have completely overlooked the speaking style despite differences found among binary and non-binary populations. In this work, we synthesise gender-ambiguous speech by combining the timbre of a male speaker with the manner of speech of a female speaker using voice morphing and pitch shifting towards the male-female boundary. Subjective evaluations indicate that the ambiguity of the morphed samples that convey the female speech style is higher than those that undergo plain pitch transformations suggesting that the speaking style can be a contributing factor in creating gender-ambiguous speech. To our knowledge, this is the first study that explicitly uses the transfer of the speaking style to create gender-ambiguous voices.
Silent Speech Interfaces (SSIs) offer a noninvasive alternative to brain-computer interfaces for soundless verbal communication. We introduce Multimodal Orofacial Neural Audio (MONA), a system that leverages cross-modal alignment through novel loss functions--cross-contrast (crossCon) and supervised temporal contrast (supTcon)--to train a multimodal model with a shared latent representation. This architecture enables the use of audio-only datasets like LibriSpeech to improve silent speech recognition. Additionally, our introduction of Large Language Model (LLM) Integrated Scoring Adjustment (LISA) significantly improves recognition accuracy. Together, MONA LISA reduces the state-of-the-art word error rate (WER) from 28.8% to 12.2% in the Gaddy (2020) benchmark dataset for silent speech on an open vocabulary. For vocal EMG recordings, our method improves the state-of-the-art from 23.3% to 3.7% WER. In the Brain-to-Text 2024 competition, LISA performs best, improving the top WER from 9.8% to 8.9%. To the best of our knowledge, this work represents the first instance where noninvasive silent speech recognition on an open vocabulary has cleared the threshold of 15% WER, demonstrating that SSIs can be a viable alternative to automatic speech recognition (ASR). Our work not only narrows the performance gap between silent and vocalized speech but also opens new possibilities in human-computer interaction, demonstrating the potential of cross-modal approaches in noisy and data-limited regimes.
Self-supervised speech representation learning enables the extraction of meaningful features from raw waveforms. These features can then be efficiently used across multiple downstream tasks. However, two significant issues arise when considering the deployment of such methods ``in-the-wild": (i) Their large size, which can be prohibitive for edge applications; and (ii) their robustness to detrimental factors, such as noise and/or reverberation, that can heavily degrade the performance of such systems. In this work, we propose RobustDistiller, a novel knowledge distillation mechanism that tackles both problems jointly. Simultaneously to the distillation recipe, we apply a multi-task learning objective to encourage the network to learn noise-invariant representations by denoising the input. The proposed mechanism is evaluated on twelve different downstream tasks. It outperforms several benchmarks regardless of noise type, or noise and reverberation levels. Experimental results show that the new Student model with 23M parameters can achieve results comparable to the Teacher model with 95M parameters. Lastly, we show that the proposed recipe can be applied to other distillation methodologies, such as the recent DPWavLM. For reproducibility, code and model checkpoints will be made available at \mbox{\url{https://github.com/Hguimaraes/robustdistiller}}.
Over the years there has been ongoing interest in detecting authorship of a text based on statistical properties of the text, such as by using occurrence rates of noncontextual words. In previous work, these techniques have been used, for example, to determine authorship of all of \emph{The Federalist Papers}. Such methods may be useful in more modern times to detect fake or AI authorship. Progress in statistical natural language parsers introduces the possibility of using grammatical structure to detect authorship. In this paper we explore a new possibility for detecting authorship using grammatical structural information extracted using a statistical natural language parser. This paper provides a proof of concept, testing author classification based on grammatical structure on a set of "proof texts," The Federalist Papers and Sanditon which have been as test cases in previous authorship detection studies. Several features extracted from the statistical natural language parser were explored: all subtrees of some depth from any level; rooted subtrees of some depth, part of speech, and part of speech by level in the parse tree. It was found to be helpful to project the features into a lower dimensional space. Statistical experiments on these documents demonstrate that information from a statistical parser can, in fact, assist in distinguishing authors.
Recent advancements in multimodal Human-Robot Interaction (HRI) datasets have highlighted the fusion of speech and gesture, expanding robots' capabilities to absorb explicit and implicit HRI insights. However, existing speech-gesture HRI datasets often focus on elementary tasks, like object pointing and pushing, revealing limitations in scaling to intricate domains and prioritizing human command data over robot behavior records. To bridge these gaps, we introduce NatSGD, a multimodal HRI dataset encompassing human commands through speech and gestures that are natural, synchronized with robot behavior demonstrations. NatSGD serves as a foundational resource at the intersection of machine learning and HRI research, and we demonstrate its effectiveness in training robots to understand tasks through multimodal human commands, emphasizing the significance of jointly considering speech and gestures. We have released our dataset, simulator, and code to facilitate future research in human-robot interaction system learning; access these resources at https://www.snehesh.com/natsgd/
Inspired by the activity-silent and persistent activity mechanisms in human visual perception biology, we design a Unified Static and Dynamic Network (UniSDNet), to learn the semantic association between the video and text/audio queries in a cross-modal environment for efficient video grounding. For static modeling, we devise a novel residual structure (ResMLP) to boost the global comprehensive interaction between the video segments and queries, achieving more effective semantic enhancement/supplement. For dynamic modeling, we effectively exploit three characteristics of the persistent activity mechanism in our network design for a better video context comprehension. Specifically, we construct a diffusely connected video clip graph on the basis of 2D sparse temporal masking to reflect the "short-term effect" relationship. We innovatively consider the temporal distance and relevance as the joint "auxiliary evidence clues" and design a multi-kernel Temporal Gaussian Filter to expand the context clue into high-dimensional space, simulating the "complex visual perception", and then conduct element level filtering convolution operations on neighbour clip nodes in message passing stage for finally generating and ranking the candidate proposals. Our UniSDNet is applicable to both Natural Language Video Grounding (NLVG) and Spoken Language Video Grounding (SLVG) tasks. Our UniSDNet achieves SOTA performance on three widely used datasets for NLVG, as well as three datasets for SLVG, e.g., reporting new records at 38.88% R@1,IoU@0.7 on ActivityNet Captions and 40.26% R@1,IoU@0.5 on TACoS. To facilitate this field, we collect two new datasets (Charades-STA Speech and TACoS Speech) for SLVG task. Meanwhile, the inference speed of our UniSDNet is 1.56$\times$ faster than the strong multi-query benchmark. Code is available at: https://github.com/xian-sh/UniSDNet.