Cem Subakan

Luca Della Libera, Cem Subakan, Mirco Ravanelli et al. · cmu

Transformers have recently achieved state-of-the-art performance in speech separation. These models, however, are computationally demanding and require a lot of learnable parameters. This paper explores Transformer-based speech separation with a reduced computational cost. Our main contribution is the development of the Resource-Efficient Separation Transformer (RE-SepFormer), a self-attention-based architecture that reduces the computational burden in two ways. First, it uses non-overlapping blocks in the latent space. Second, it operates on compact latent summaries calculated from each chunk. The RE-SepFormer reaches a competitive performance on the popular WSJ0-2Mix and WHAM! datasets in both causal and non-causal settings. Remarkably, it scales significantly better than the previous Transformer-based architectures in terms of memory and inference time, making it more suitable for processing long mixtures.

Luca Della Libera, Pooneh Mousavi, Salah Zaiem et al.

Modern multilingual automatic speech recognition (ASR) systems like Whisper have made it possible to transcribe audio in multiple languages with a single model. However, current state-of-the-art ASR models are typically evaluated on individual languages or in a multi-task setting, overlooking the challenge of continually learning new languages. There is insufficient research on how to add new languages without losing valuable information from previous data. Furthermore, existing continual learning benchmarks focus mostly on vision and language tasks, leaving continual learning for multilingual ASR largely unexplored. To bridge this gap, we propose CL-MASR, a benchmark designed for studying multilingual ASR in a continual learning setting. CL-MASR provides a diverse set of continual learning methods implemented on top of large-scale pretrained ASR models, along with common metrics to assess the effectiveness of learning new languages while addressing the issue of catastrophic forgetting. To the best of our knowledge, CL-MASR is the first continual learning benchmark for the multilingual ASR task. The code is available at https://github.com/speechbrain/benchmarks.

Zhepei Wang, Cem Subakan, Xilin Jiang et al. · uw

In this paper, we work on a sound recognition system that continually incorporates new sound classes. Our main goal is to develop a framework where the model can be updated without relying on labeled data. For this purpose, we propose adopting representation learning, where an encoder is trained using unlabeled data. This learning framework enables the study and implementation of a practically relevant use case where only a small amount of the labels is available in a continual learning context. We also make the empirical observation that a similarity-based representation learning method within this framework is robust to forgetting even if no explicit mechanism against forgetting is employed. We show that this approach obtains similar performance compared to several distillation-based continual learning methods when employed on self-supervised representation learning methods.

Luca Della Libera, Cem Subakan, Mirco Ravanelli

Neural audio codecs are at the core of modern conversational speech technologies, converting continuous speech into sequences of discrete tokens that can be processed by LLMs. However, existing codecs typically operate at fixed frame rates, allocating tokens uniformly in time and producing unnecessarily long sequences. In this work, we introduce DyCAST, a Dynamic Character-Aligned Speech Tokenizer that enables variable-frame-rate tokenization through soft character-level alignment and explicit duration modeling. DyCAST learns to associate tokens with character-level linguistic units during training and supports alignment-free inference with direct control over token durations at decoding time. To improve speech resynthesis quality at low frame rates, we further introduce a retrieval-augmented decoding mechanism that enhances reconstruction fidelity without increasing bitrate. Experiments show that DyCAST achieves competitive speech resynthesis quality and downstream performance while using significantly fewer tokens than fixed-frame-rate codecs. Code and checkpoints will be released publicly at https://github.com/lucadellalib/dycast.

Francesco Paissan, Luca Della Libera, Zhepei Wang et al.

In this paper, we explore audio-editing with non-rigid text edits. We show that the proposed editing pipeline is able to create audio edits that remain faithful to the input audio. We explore text prompts that perform addition, style transfer, and in-painting. We quantitatively and qualitatively show that the edits are able to obtain results which outperform Audio-LDM, a recently released text-prompted audio generation model. Qualitative inspection of the results points out that the edits given by our approach remain more faithful to the input audio in terms of keeping the original onsets and offsets of the audio events.

Francesco Paissan, Cem Subakan, Mirco Ravanelli

In this paper, we introduce a new approach, called Posthoc Interpretation via Quantization (PIQ), for interpreting decisions made by trained classifiers. Our method utilizes vector quantization to transform the representations of a classifier into a discrete, class-specific latent space. The class-specific codebooks act as a bottleneck that forces the interpreter to focus on the parts of the input data deemed relevant by the classifier for making a prediction. Our model formulation also enables learning concepts by incorporating the supervision of pretrained annotation models such as state-of-the-art image segmentation models. We evaluated our method through quantitative and qualitative studies involving black-and-white images, color images, and audio. As a result of these studies we found that PIQ generates interpretations that are more easily understood by participants to our user studies when compared to several other interpretation methods in the literature.

Eleonora Mancini, Francesco Paissan, Mirco Ravanelli et al.

Neural networks are typically black-boxes that remain opaque with regards to their decision mechanisms. Several works in the literature have proposed post-hoc explanation methods to alleviate this issue. This paper proposes LMAC-TD, a post-hoc explanation method that trains a decoder to produce explanations directly in the time domain. This methodology builds upon the foundation of L-MAC, Listenable Maps for Audio Classifiers, a method that produces faithful and listenable explanations. We incorporate SepFormer, a popular transformer-based time-domain source separation architecture. We show through a user study that LMAC-TD significantly improves the audio quality of the produced explanations while not sacrificing from faithfulness.

Jihoon Jeong, Pooneh Mousavi, Mirco Ravanelli et al.

Large audio-language models (LALMs) can generate reasoning chains for their predictions, but it remains unclear whether these reasoning chains remain grounded in the input audio. In this paper, we propose an RL-based strategy that grounds the reasoning outputs of LALMs with explicit timestamp annotations referring to relevant segments of the audio signal. Our analysis shows that timestamp grounding leads the model to attend more strongly to audio tokens during reasoning generation. Experiments on four speech-based benchmark datasets demonstrate that our approach improves performance compared to both zero-shot reasoning and fine-tuning without timestamp grounding. Additionally, grounding amplifies desirable reasoning behaviors, such as region exploration, audiology verification, and consistency, underscoring the importance of grounding mechanisms for faithful multimodal reasoning.

Jingyi Li, Luca Della Libera, Mirco Ravanelli et al.

Many speech enhancement (SE) methods rely on continuous representations. Recently, discrete audio tokens have been explored to enable autoregressive generation for SE. However, it remains unclear whether discretization itself consistently improves SE performance. In this paper, we introduce LL-SDR, a token-based speech enhancement framework that explicitly leverages discretization to better separate speech and noise. Our first contribution is a Variance-Ordered Residual Vector Quantizer (VO-RVQ), designed to disentangle speech and noise distributions during tokenization. Second, we propose a latent-space discriminator to better align enhanced embeddings with semantic embeddings. Experiments show that LL-SDR outperforms continuous baselines and matches the performance of autoregressive token-based approaches, while enabling lightweight, low-latency speech enhancement in both reverberant and non-reverberant noisy environments. Demos and source code are available at our project websites.

Francesco Paissan, Luca Della Libera, Mirco Ravanelli et al.

Neural audio codecs provide compact discrete representations for speech generation and manipulation. However, most codecs organize tokens as frame-level sequences, making it difficult to study or intervene on global factors of variation. In this work, we propose the Latent Audio Tokenizer for Token-space Editing (LATTE) that appends a fixed set of learnable latent tokens to the audio feature sequence and retains only these tokens for quantization and decoding. This design produces a compact, non-temporally aligned bottleneck in which each token can aggregate global information across the full utterance. We show that the resulting tokenizer preserves competitive reconstruction quality in low-bitrate speech coding settings while enabling simple token-space interventions. In particular, we find that swapping selected latent token positions between utterances can modify global attributes, such as speaker identity and background noise, and we evaluate these interventions on voice conversion and denoising tasks. Our results suggest that compact latent audio tokenizers can support controllable audio manipulation without supervision in task-specific editing models.

Dinh Phu Tran, Jihoon Jeong, Saad Wazir et al.

We present a formal problem formulation for \textit{Reliable} Audio-Visual Question Answering ($\mathcal{R}$-AVQA), where we prefer abstention over answering incorrectly. While recent AVQA models have high accuracy, their ability to identify when they are likely wrong and their consequent abstention from answering remain underexplored areas of research. To fill this gap, we explore several approaches and then propose Adaptive Confidence Refinement (ACR), a lightweight method to further enhance the performance of $\mathcal{R}$-AVQA. Our key insight is that the Maximum Softmax Probability (MSP) is Bayes-optimal only under strong calibration, a condition usually not met in deep neural networks, particularly in multimodal models. Instead of replacing MSP, our ACR maintains it as a primary confidence signal and applies input-adaptive residual corrections when MSP is deemed unreliable. ACR introduces two learned heads: i) a Residual Risk Head that predicts low-magnitude correctness residuals that MSP does not capture, and ii) a Confidence Gating Head to determine MSP trustworthiness. Our experiments and theoretical analysis show that ACR consistently outperforms existing methods on in- and out-of-disrtibution, and data bias settings across three different AVQA architectures, establishing a solid foundation for $\mathcal{R}$-AVQA task. The code and checkpoints will be available upon acceptance \href{https://github.com/PhuTran1005/R-AVQA}{at here}

Luca Della Libera, Cem Subakan, Mirco Ravanelli

Neural audio codecs are a fundamental component of modern generative audio pipelines. Although recent codecs achieve strong low-bitrate reconstruction and provide powerful representations for downstream tasks, most are non-streamable, limiting their use in real-time applications. We present FocalCodec-Stream, a hybrid codec based on focal modulation that compresses speech into a single binary codebook at 0.55 - 0.80 kbps with a theoretical latency of 80 ms. Our approach combines multi-stage causal distillation of WavLM with targeted architectural improvements, including a lightweight refiner module that enhances quality under latency constraints. Experiments show that FocalCodec-Stream outperforms existing streamable codecs at comparable bitrates, while preserving both semantic and acoustic information. The result is a favorable trade-off between reconstruction quality, downstream task performance, latency, and efficiency. Code and checkpoints will be released at https://github.com/lucadellalib/focalcodec.

Mirco Ravanelli, Titouan Parcollet, Adel Moumen et al.

SpeechBrain is an open-source Conversational AI toolkit based on PyTorch, focused particularly on speech processing tasks such as speech recognition, speech enhancement, speaker recognition, text-to-speech, and much more. It promotes transparency and replicability by releasing both the pre-trained models and the complete "recipes" of code and algorithms required for training them. This paper presents SpeechBrain 1.0, a significant milestone in the evolution of the toolkit, which now has over 200 recipes for speech, audio, and language processing tasks, and more than 100 models available on Hugging Face. SpeechBrain 1.0 introduces new technologies to support diverse learning modalities, Large Language Model (LLM) integration, and advanced decoding strategies, along with novel models, tasks, and modalities. It also includes a new benchmark repository, offering researchers a unified platform for evaluating models across diverse tasks.

Juan Zuluaga-Gomez, Sara Ahmed, Danielius Visockas et al.

Despite the recent advancements in Automatic Speech Recognition (ASR), the recognition of accented speech still remains a dominant problem. In order to create more inclusive ASR systems, research has shown that the integration of accent information, as part of a larger ASR framework, can lead to the mitigation of accented speech errors. We address multilingual accent classification through the ECAPA-TDNN and Wav2Vec 2.0/XLSR architectures which have been proven to perform well on a variety of speech-related downstream tasks. We introduce a simple-to-follow recipe aligned to the SpeechBrain toolkit for accent classification based on Common Voice 7.0 (English) and Common Voice 11.0 (Italian, German, and Spanish). Furthermore, we establish new state-of-the-art for English accent classification with as high as 95% accuracy. We also study the internal categorization of the Wav2Vev 2.0 embeddings through t-SNE, noting that there is a level of clustering based on phonological similarity. (Our recipe is open-source in the SpeechBrain toolkit, see: https://github.com/speechbrain/speechbrain/tree/develop/recipes)

Mirco Ravanelli, Titouan Parcollet, Peter Plantinga et al.

SpeechBrain is an open-source and all-in-one speech toolkit. It is designed to facilitate the research and development of neural speech processing technologies by being simple, flexible, user-friendly, and well-documented. This paper describes the core architecture designed to support several tasks of common interest, allowing users to naturally conceive, compare and share novel speech processing pipelines. SpeechBrain achieves competitive or state-of-the-art performance in a wide range of speech benchmarks. It also provides training recipes, pretrained models, and inference scripts for popular speech datasets, as well as tutorials which allow anyone with basic Python proficiency to familiarize themselves with speech technologies.

Pooneh Mousavi, Gallil Maimon, Adel Moumen et al.

Discrete audio tokens are compact representations that aim to preserve perceptual quality, phonetic content, and speaker characteristics while enabling efficient storage and inference, as well as competitive performance across diverse downstream tasks. They provide a practical alternative to continuous features, enabling the integration of speech and audio into modern large language models (LLMs). As interest in token-based audio processing grows, various tokenization methods have emerged, and several surveys have reviewed the latest progress in the field. However, existing studies often focus on specific domains or tasks and lack a unified comparison across various benchmarks. This paper presents a systematic review and benchmark of discrete audio tokenizers, covering three domains: speech, music, and general audio. We propose a taxonomy of tokenization approaches based on encoder-decoder, quantization techniques, training paradigm, streamability, and application domains. We evaluate tokenizers on multiple benchmarks for reconstruction, downstream performance, and acoustic language modeling, and analyze trade-offs through controlled ablation studies. Our findings highlight key limitations, practical considerations, and open challenges, providing insight and guidance for future research in this rapidly evolving area. For more information, including our main results and tokenizer database, please refer to our website: https://poonehmousavi.github.io/dates-website/.

Luca Della Libera, Francesco Paissan, Cem Subakan et al.

Large language models have revolutionized natural language processing through self-supervised pretraining on massive datasets. Inspired by this success, researchers have explored adapting these methods to speech by discretizing continuous audio into tokens using neural audio codecs. However, existing approaches face limitations, including high bitrates, the loss of either semantic or acoustic information, and the reliance on multi-codebook designs when trying to capture both, which increases architectural complexity for downstream tasks. To address these challenges, we introduce FocalCodec, an efficient low-bitrate codec based on focal modulation that utilizes a single binary codebook to compress speech between 0.16 and 0.65 kbps. FocalCodec delivers competitive performance in speech resynthesis and voice conversion at lower bitrates than the current state-of-the-art, while effectively handling multilingual speech and noisy environments. Evaluation on downstream tasks shows that FocalCodec successfully preserves sufficient semantic and acoustic information, while also being well-suited for generative modeling. Demo samples and code are available at https://lucadellalib.github.io/focalcodec-web/.

Eleonora Mancini, Francesco Paissan, Paolo Torroni et al.

Speech impairments in Parkinson's disease (PD) provide significant early indicators for diagnosis. While models for speech-based PD detection have shown strong performance, their interpretability remains underexplored. This study systematically evaluates several explainability methods to identify PD-specific speech features, aiming to support the development of accurate, interpretable models for clinical decision-making in PD diagnosis and monitoring. Our methodology involves (i) obtaining attributions and saliency maps using mainstream interpretability techniques, (ii) quantitatively evaluating the faithfulness of these maps and their combinations obtained via union and intersection through a range of established metrics, and (iii) assessing the information conveyed by the saliency maps for PD detection from an auxiliary classifier. Our results reveal that, while explanations are aligned with the classifier, they often fail to provide valuable information for domain experts.

Luca Della Libera, Cem Subakan, Mirco Ravanelli

The increasing success of deep neural networks has raised concerns about their inherent black-box nature, posing challenges related to interpretability and trust. While there has been extensive exploration of interpretation techniques in vision and language, interpretability in the audio domain has received limited attention, primarily focusing on post-hoc explanations. This paper addresses the problem of interpretability by-design in the audio domain by utilizing the recently proposed attention-free focal modulation networks (FocalNets). We apply FocalNets to the task of environmental sound classification for the first time and evaluate their interpretability properties on the popular ESC-50 dataset. Our method outperforms a similarly sized vision transformer both in terms of accuracy and interpretability. Furthermore, it is competitive against PIQ, a method specifically designed for post-hoc interpretation in the audio domain.

Luca Della Libera, Cem Subakan, Mirco Ravanelli

In speech processing pipelines, improving the quality and intelligibility of real-world recordings is crucial. While supervised regression is the primary method for speech enhancement, audio tokenization is emerging as a promising alternative for a smooth integration with other modalities. However, research on speech enhancement using discrete representations is still limited. Previous work has mainly focused on semantic tokens, which tend to discard key acoustic details such as speaker identity. Additionally, these studies typically employ non-autoregressive models, assuming conditional independence of outputs and overlooking the potential improvements offered by autoregressive modeling. To address these gaps we: 1) conduct a comprehensive study of the performance of acoustic tokens for speech enhancement, including the effect of bitrate and noise strength; 2) introduce a novel transducer-based autoregressive architecture specifically designed for this task. Experiments on VoiceBank and Libri1Mix datasets show that acoustic tokens outperform semantic tokens in terms of preserving speaker identity, and that our autoregressive approach can further improve performance. Nevertheless, we observe that discrete representations still fall short compared to continuous ones, highlighting the need for further research in this area.

Pooneh Mousavi, Shubham Gupta, Cem Subakan et al.

Foundation models based on large language models (LLMs) have shown great success in handling various tasks and modalities. However, adapting these models for general-purpose audio-language tasks is challenging due to differences in acoustic environments and task variations. In this work, we introduce LiSTEN Learning Soft Token Embeddings for Neural Audio LLMs), a framework for adapting LLMs to speech and audio tasks. LiSTEN uses a dynamic prompt selection strategy with learnable key-value pairs, allowing the model to balance general and task-specific knowledge while avoiding overfitting in a multitask setting. Our approach reduces dependence on large-scale ASR or captioning datasets, achieves competitive performance with fewer trainable parameters, and simplifies training by using a single-stage process. Additionally, LiSTEN enhances interpretability by analyzing the diversity and overlap of selected prompts across different tasks.

Anthony Deschênes, Rémi Georges, Cem Subakan et al.

In recent years, the wood product industry has been facing a skilled labor shortage. The result is more frequent sudden failures, resulting in additional costs for these companies already operating in a very competitive market. Moreover, sawmills are challenging environments for machinery and sensors. Given that experienced machine operators may be able to diagnose defects or malfunctions, one possible way of assisting novice operators is through acoustic monitoring. As a step towards the automation of wood-processing equipment and decision support systems for machine operators, in this paper, we explore using a deep convolutional autoencoder for acoustic anomaly detection of wood planers on a new real-life dataset. Specifically, our convolutional autoencoder with skip connections (Skip-CAE) and our Skip-CAE transformer outperform the DCASE autoencoder baseline, one-class SVM, isolation forest and a published convolutional autoencoder architecture, respectively obtaining an area under the ROC curve of 0.846 and 0.875 on a dataset of real-factory planer sounds. Moreover, we show that adding skip connections and attention mechanism under the form of a transformer encoder-decoder helps to further improve the anomaly detection capabilities.

Maab Elrashid, Anthony Deschênes, Cem Subakan et al.

Interpretability is essential for user trust in real-world anomaly detection applications. However, deep learning models, despite their strong performance, often lack transparency. In this work, we study the interpretability of autoencoder-based models for audio anomaly detection, by comparing a standard autoencoder (AE) with a mask autoencoder (MAE) in terms of detection performance and interpretability. We applied several attribution methods, including error maps, saliency maps, SmoothGrad, Integrated Gradients, GradSHAP, and Grad-CAM. Although MAE shows a slightly lower detection, it consistently provides more faithful and temporally precise explanations, suggesting a better alignment with true anomalies. To assess the relevance of the regions highlighted by the explanation method, we propose a perturbation-based faithfulness metric that replaces them with their reconstructions to simulate normal input. Our findings, based on experiments in a real industrial scenario, highlight the importance of incorporating interpretability into anomaly detection pipelines and show that masked training improves explanation quality without compromising performance.

Luca Della Libera, Cem Subakan, Mirco Ravanelli

Large language models show that simple autoregressive training can yield scalable and coherent generation, but extending this paradigm to speech remains challenging due to the entanglement of semantic and acoustic information. Most existing speech language models rely on text supervision, hierarchical token streams, or complex hybrid architectures, departing from the single-stream generative pretraining paradigm that has proven effective in text. In this work, we introduce WavSLM, a speech language model trained by quantizing and distilling self-supervised WavLM representations into a single codebook and optimizing an autoregressive next-chunk prediction objective. WavSLM jointly models semantic and acoustic information within a single token stream without text supervision or text pretraining. Despite its simplicity, it achieves competitive performance on consistency benchmarks and speech generation while using fewer parameters, less training data, and supporting streaming inference. Demo samples are available at https://lucadellalib.github.io/wavslm-web/.

Lovenya Jain, Pooneh Mousavi, Mirco Ravanelli et al.

Faithfulness measures whether chain-of-thought (CoT) representations accurately reflect a model's decision process and can be used as reliable explanations. Prior work has shown that CoTs from text-based LLMs are often unfaithful. This question has not been explored for large audio-language models (LALMs), where faithfulness is critical for safety-sensitive applications. Reasoning in LALMs is also more challenging, as models must first extract relevant clues from audio before reasoning over them. In this paper, we investigate the faithfulness of CoTs produced by several LALMs by applying targeted interventions, including paraphrasing, filler token injection, early answering, and introducing mistakes, on two challenging reasoning datasets: SAKURA and MMAR. After going through the aforementioned interventions across several datasets and tasks, our experiments suggest that, LALMs generally produce CoTs that appear to be faithful to their underlying decision processes.

Matthieu Cervera, Francesco Paissan, Mirco Ravanelli et al.

Free-form, text-based audio editing remains a persistent challenge, despite progress in inversion-based neural methods. Current approaches rely on slow inversion procedures, limiting their practicality. We present a virtual-consistency based audio editing system that bypasses inversion by adapting the sampling process of diffusion models. Our pipeline is model-agnostic, requiring no fine-tuning or architectural changes, and achieves substantial speed-ups over recent neural editing baselines. Crucially, it achieves this efficiency without compromising quality, as demonstrated by quantitative benchmarks and a user study involving 16 participants.

Mathis Rezzouk, Fabrice Gagnon, Alyson Champagne et al.

EEG-based analysis of pain perception, enhanced by machine learning, reveals how the brain encodes pain by identifying neural patterns evoked by noxious stimulation. However, a major challenge that remains is the generalization of machine learning models across individuals, given the high cross-participant variability inherent to EEG signals and the limited focus on direct pain perception identification in current research. In this study, we systematically evaluate the performance of cross-participant generalization of a wide range of models, including traditional classifiers and deep neural classifiers for identifying the sensory modality of thermal pain and aversive auditory stimulation from EEG recordings. Using a novel dataset of EEG recordings from 108 participants, we benchmark model performance under both within- and cross-participant evaluation settings. Our findings show that traditional models suffered the largest drop from within- to cross-participant performance, while deep learning models proved more resilient, underscoring their potential for subject-invariant EEG decoding. Even though performance variability remained high, the strong results of the graph-based model highlight its potential to capture subject-invariant structure in EEG signals. On the other hand, we also share the preprocessed dataset used in this study, providing a standardized benchmark for evaluating future algorithms under the same generalization constraints.

Jacob Comeau, Mathieu Bazinet, Pascal Germain et al.

Continual learning algorithms aim to learn from a sequence of tasks, making the training distribution non-stationary. The majority of existing continual learning approaches in the literature rely on heuristics and do not provide learning guarantees. In this paper, we present a new method called Continual Pick-to-Learn (CoP2L), which is able to retain the most representative samples for each task in an efficient way. CoP2L combines the Pick-to-Learn algorithm (rooted in the sample compression theory) and the experience replay continual learning scheme. This allows us to provide non-vacuous upper bounds on the generalization loss of the learned predictors, numerically computable after each task. We empirically evaluate our approach on several standard continual learning benchmarks across Class-Incremental, Task-Incremental, and Domain-Incremental settings. Our results show that CoP2L is highly competitive across all setups, often outperforming existing baselines, and significantly mitigating catastrophic forgetting compared to vanilla experience replay in the Class-Incremental setting. It is possible to leverage the bounds provided by CoP2L in practical scenarios to certify the predictor reliability on previously learned tasks, in order to improve the trustworthiness of the continual learning algorithm.

Shubham Gupta, Zichao Li, Tianyi Chen et al.

Document retrieval is a core component of question-answering systems, as it enables conditioning answer generation on new and large-scale corpora. While effective, the standard practice of encoding documents into high-dimensional embeddings for similarity search entails large memory and compute footprints, and also makes it hard to inspect the inner workings of the system. In this paper, we propose a tree-based method for organizing and representing reference documents at various granular levels, which offers the flexibility to balance cost and utility, and eases the inspection of the corpus content and retrieval operations. Our method, called ReTreever, jointly learns a routing function per internal node of a binary tree such that query and reference documents are assigned to similar tree branches, hence directly optimizing for retrieval performance. Our evaluations show that ReTreever generally preserves full representation accuracy. Its hierarchical structure further provides strong coarse representations and enhances transparency by indirectly learning meaningful semantic groupings. Among hierarchical retrieval methods, ReTreever achieves the best retrieval accuracy at the lowest latency, proving that this family of techniques can be viable in practical applications.

Pooneh Mousavi, Luca Della Libera, Jarod Duret et al.

Discrete audio tokens have recently gained considerable attention for their potential to connect audio and language processing, enabling the creation of modern multimodal large language models. Ideal audio tokens must effectively preserve phonetic and semantic content along with paralinguistic information, speaker identity, and other details. While several types of audio tokens have been recently proposed, identifying the optimal tokenizer for various tasks is challenging due to the inconsistent evaluation settings in existing studies. To address this gap, we release the Discrete Audio and Speech Benchmark (DASB), a comprehensive leaderboard for benchmarking discrete audio tokens across a wide range of discriminative tasks, including speech recognition, speaker identification and verification, emotion recognition, keyword spotting, and intent classification, as well as generative tasks such as speech enhancement, separation, and text-to-speech. Our results show that, on average, semantic tokens outperform compression tokens across most discriminative and generative tasks. However, the performance gap between semantic tokens and standard continuous representations remains substantial, highlighting the need for further research in this field.

Pooneh Mousavi, Jarod Duret, Salah Zaiem et al.

Discrete audio tokens have recently gained attention for their potential to bridge the gap between audio and language processing. Ideal audio tokens must preserve content, paralinguistic elements, speaker identity, and many other audio details. Current audio tokenization methods fall into two categories: Semantic tokens, acquired through quantization of Self-Supervised Learning (SSL) models, and Neural compression-based tokens (codecs). Although previous studies have benchmarked codec models to identify optimal configurations, the ideal setup for quantizing pretrained SSL models remains unclear. This paper explores the optimal configuration of semantic tokens across discriminative and generative tasks. We propose a scalable solution to train a universal vocoder across multiple SSL layers. Furthermore, an attention mechanism is employed to identify task-specific influential layers, enhancing the adaptability and performance of semantic tokens in diverse audio applications.

Francesco Paissan, Mirco Ravanelli, Cem Subakan

Despite the impressive performance of deep learning models across diverse tasks, their complexity poses challenges for interpretation. This challenge is particularly evident for audio signals, where conveying interpretations becomes inherently difficult. To address this issue, we introduce Listenable Maps for Audio Classifiers (L-MAC), a posthoc interpretation method that generates faithful and listenable interpretations. L-MAC utilizes a decoder on top of a pretrained classifier to generate binary masks that highlight relevant portions of the input audio. We train the decoder with a loss function that maximizes the confidence of the classifier decision on the masked-in portion of the audio while minimizing the probability of model output for the masked-out portion. Quantitative evaluations on both in-domain and out-of-domain data demonstrate that L-MAC consistently produces more faithful interpretations than several gradient and masking-based methodologies. Furthermore, a user study confirms that, on average, users prefer the interpretations generated by the proposed technique.

Zhepei Wang, Cem Subakan, Krishna Subramani et al.

Recent advances in using language models to obtain cross-modal audio-text representations have overcome the limitations of conventional training approaches that use predefined labels. This has allowed the community to make progress in tasks like zero-shot classification, which would otherwise not be possible. However, learning such representations requires a large amount of human-annotated audio-text pairs. In this paper, we study unsupervised approaches to improve the learning framework of such representations with unpaired text and audio. We explore domain-unspecific and domain-specific curation methods to create audio-text pairs that we use to further improve the model. We also show that when domain-specific curation is used in conjunction with a soft-labeled contrastive loss, we are able to obtain significant improvement in terms of zero-shot classification performance on downstream sound event classification or acoustic scene classification tasks.

Arsenii Gorin, Cem Subakan, Sajjad Abdoli et al.

In this paper, we explore self-supervised learning (SSL) for analyzing a first-of-its-kind database of cry recordings containing clinical indications of more than a thousand newborns. Specifically, we target cry-based detection of neurological injury as well as identification of cry triggers such as pain, hunger, and discomfort. Annotating a large database in the medical setting is expensive and time-consuming, typically requiring the collaboration of several experts over years. Leveraging large amounts of unlabeled audio data to learn useful representations can lower the cost of building robust models and, ultimately, clinical solutions. In this work, we experiment with self-supervised pre-training of a convolutional neural network on large audio datasets. We show that pre-training with SSL contrastive loss (SimCLR) performs significantly better than supervised pre-training for both neuro injury and cry triggers. In addition, we demonstrate further performance gains through SSL-based domain adaptation using unlabeled infant cries. We also show that using such SSL-based pre-training for adaptation to cry sounds decreases the need for labeled data of the overall system.

David Budaghyan, Charles C. Onu, Arsenii Gorin et al.

This paper describes the Ubenwa CryCeleb dataset - a labeled collection of infant cries - and the accompanying CryCeleb 2023 task, which is a public speaker verification challenge based on cry sounds. We released more than 6 hours of manually segmented cry sounds from 786 newborns for academic use, aiming to encourage research in infant cry analysis. The inaugural public competition attracted 59 participants, 11 of whom improved the baseline performance. The top-performing system achieved a significant improvement scoring 25.8% equal error rate, which is still far from the performance of state-of-the-art adult speaker verification systems. Therefore, we believe there is room for further research on this dataset, potentially extending beyond the verification task.

Cem Subakan, Mirco Ravanelli, Samuele Cornell et al.

Transformers have enabled impressive improvements in deep learning. They often outperform recurrent and convolutional models in many tasks while taking advantage of parallel processing. Recently, we proposed the SepFormer, which obtains state-of-the-art performance in speech separation with the WSJ0-2/3 Mix datasets. This paper studies in-depth Transformers for speech separation. In particular, we extend our previous findings on the SepFormer by providing results on more challenging noisy and noisy-reverberant datasets, such as LibriMix, WHAM!, and WHAMR!. Moreover, we extend our model to perform speech enhancement and provide experimental evidence on denoising and dereverberation tasks. Finally, we investigate, for the first time in speech separation, the use of efficient self-attention mechanisms such as Linformers, Lonformers, and ReFormers. We found that they reduce memory requirements significantly. For example, we show that the Reformer-based attention outperforms the popular Conv-TasNet model on the WSJ0-2Mix dataset while being faster at inference and comparable in terms of memory consumption.

Cem Subakan, Mirco Ravanelli, Samuele Cornell et al.

In recent years, deep learning based source separation has achieved impressive results. Most studies, however, still evaluate separation models on synthetic datasets, while the performance of state-of-the-art techniques on in-the-wild speech data remains an open question. This paper contributes to fill this gap in two ways. First, we release the REAL-M dataset, a crowd-sourced corpus of real-life mixtures. Secondly, we address the problem of performance evaluation of real-life mixtures, where the ground truth is not available. We bypass this issue by carefully designing a blind Scale-Invariant Signal-to-Noise Ratio (SI-SNR) neural estimator. Through a user study, we show that our estimator reliably evaluates the separation performance on real mixtures. The performance predictions of the SI-SNR estimator indeed correlate well with human opinions. Moreover, we observe that the performance trends predicted by our estimator on the REAL-M dataset closely follow those achieved on synthetic benchmarks when evaluating popular speech separation models.

Cem Subakan, Mirco Ravanelli, Samuele Cornell et al.

Recurrent Neural Networks (RNNs) have long been the dominant architecture in sequence-to-sequence learning. RNNs, however, are inherently sequential models that do not allow parallelization of their computations. Transformers are emerging as a natural alternative to standard RNNs, replacing recurrent computations with a multi-head attention mechanism. In this paper, we propose the SepFormer, a novel RNN-free Transformer-based neural network for speech separation. The SepFormer learns short and long-term dependencies with a multi-scale approach that employs transformers. The proposed model achieves state-of-the-art (SOTA) performance on the standard WSJ0-2/3mix datasets. It reaches an SI-SNRi of 22.3 dB on WSJ0-2mix and an SI-SNRi of 19.5 dB on WSJ0-3mix. The SepFormer inherits the parallelization advantages of Transformers and achieves a competitive performance even when downsampling the encoded representation by a factor of 8. It is thus significantly faster and it is less memory-demanding than the latest speech separation systems with comparable performance.

Efthymios Tzinis, Shrikant Venkataramani, Zhepei Wang et al.

In this paper, we propose a two-step training procedure for source separation via a deep neural network. In the first step we learn a transform (and it's inverse) to a latent space where masking-based separation performance using oracles is optimal. For the second step, we train a separation module that operates on the previously learned space. In order to do so, we also make use of a scale-invariant signal to distortion ratio (SI-SDR) loss function that works in the latent space, and we prove that it lower-bounds the SI-SDR in the time domain. We run various sound separation experiments that show how this approach can obtain better performance as compared to systems that learn the transform and the separation module jointly. The proposed methodology is general enough to be applicable to a large class of neural network end-to-end separation systems.

Zhepei Wang, Cem Subakan, Efthymios Tzinis et al.

Continual learning consists in incrementally training a model on a sequence of datasets and testing on the union of all datasets. In this paper, we examine continual learning for the problem of sound classification, in which we wish to refine already trained models to learn new sound classes. In practice one does not want to maintain all past training data and retrain from scratch, but naively updating a model with new data(sets) results in a degradation of already learned tasks, which is referred to as "catastrophic forgetting." We develop a generative replay procedure for generating training audio spectrogram data, in place of keeping older training datasets. We show that by incrementally refining a classifier with generative replay a generator that is 4% of the size of all previous training data matches the performance of refining the classifier keeping 20% of all previous training data. We thus conclude that we can extend a trained sound classifier to learn new classes without having to keep previously used datasets.

Cem Subakan, Oluwasanmi Koyejo, Paris Smaragdis

Popular generative model learning methods such as Generative Adversarial Networks (GANs), and Variational Autoencoders (VAE) enforce the latent representation to follow simple distributions such as isotropic Gaussian. In this paper, we argue that learning a complicated distribution over the latent space of an auto-encoder enables more accurate modeling of complicated data distributions. Based on this observation, we propose a two stage optimization procedure which maximizes an approximate implicit density model. We experimentally verify that our method outperforms GANs and VAEs on two image datasets (MNIST, CELEB-A). We also show that our approach is amenable to learning generative model for sequential data, by learning to generate speech and music.

Cem Subakan, Paris Smaragdis

Generative source separation methods such as non-negative matrix factorization (NMF) or auto-encoders, rely on the assumption of an output probability density. Generative Adversarial Networks (GANs) can learn data distributions without needing a parametric assumption on the output density. We show on a speech source separation experiment that, a multi-layer perceptron trained with a Wasserstein-GAN formulation outperforms NMF, auto-encoders trained with maximum likelihood, and variational auto-encoders in terms of source to distortion ratio.