Aalborg University
Abstract:With the advent of new sequence models like Mamba and xLSTM, several studies have shown that these models match or outperform state-of-the-art models in single-channel speech enhancement, automatic speech recognition, and self-supervised audio representation learning. However, prior research has demonstrated that sequence models like LSTM and Mamba tend to overfit to the training set. To address this issue, previous works have shown that adding self-attention to LSTMs substantially improves generalization performance for single-channel speech enhancement. Nevertheless, neither the concept of hybrid Mamba and time-frequency attention models nor their generalization performance have been explored for speech enhancement. In this paper, we propose a novel hybrid architecture, MambAttention, which combines Mamba and shared time- and frequency-multi-head attention modules for generalizable single-channel speech enhancement. To train our model, we introduce VoiceBank+Demand Extended (VB-DemandEx), a dataset inspired by VoiceBank+Demand but with more challenging noise types and lower signal-to-noise ratios. Trained on VB-DemandEx, our proposed MambAttention model significantly outperforms existing state-of-the-art LSTM-, xLSTM-, Mamba-, and Conformer-based systems of similar complexity across all reported metrics on two out-of-domain datasets: DNS 2020 and EARS-WHAM_v2, while matching their performance on the in-domain dataset VB-DemandEx. Ablation studies highlight the role of weight sharing between the time- and frequency-multi-head attention modules for generalization performance. Finally, we explore integrating the shared time- and frequency-multi-head attention modules with LSTM and xLSTM, which yields a notable performance improvement on the out-of-domain datasets. However, our MambAttention model remains superior on both out-of-domain datasets across all reported evaluation metrics.
Abstract:Recent advancements in deep learning have significantly impacted the field of speech signal processing, particularly in the analysis and manipulation of complex spectrograms. This survey provides a comprehensive overview of the state-of-the-art techniques leveraging deep neural networks for processing complex spectrograms, which encapsulate both magnitude and phase information. We begin by introducing complex spectrograms and their associated features for various speech processing tasks. Next, we explore the key components and architectures of complex-valued neural networks, which are specifically designed to handle complex-valued data and have been applied for complex spectrogram processing. We then discuss various training strategies and loss functions tailored for training neural networks to process and model complex spectrograms. The survey further examines key applications, including phase retrieval, speech enhancement, and speech separation, where deep learning has achieved significant progress by leveraging complex spectrograms or their derived feature representations. Additionally, we examine the intersection of complex spectrograms with generative models. This survey aims to serve as a valuable resource for researchers and practitioners in the field of speech signal processing and complex-valued neural networks.
Abstract:When detecting anomalous sounds in complex environments, one of the main difficulties is that trained models must be sensitive to subtle differences in monitored target signals, while many practical applications also require them to be insensitive to changes in acoustic domains. Examples of such domain shifts include changing the type of microphone or the location of acoustic sensors, which can have a much stronger impact on the acoustic signal than subtle anomalies themselves. Moreover, users typically aim to train a model only on source domain data, which they may have a relatively large collection of, and they hope that such a trained model will be able to generalize well to an unseen target domain by providing only a minimal number of samples to characterize the acoustic signals in that domain. In this work, we review and discuss recent publications focusing on this domain generalization problem for anomalous sound detection in the context of the DCASE challenges on acoustic machine condition monitoring.
Abstract:While attention-based architectures, such as Conformers, excel in speech enhancement, they face challenges such as scalability with respect to input sequence length. In contrast, the recently proposed Extended Long Short-Term Memory (xLSTM) architecture offers linear scalability. However, xLSTM-based models remain unexplored for speech enhancement. This paper introduces xLSTM-SENet, the first xLSTM-based single-channel speech enhancement system. A comparative analysis reveals that xLSTM-and notably, even LSTM-can match or outperform state-of-the-art Mamba- and Conformer-based systems across various model sizes in speech enhancement on the VoiceBank+Demand dataset. Through ablation studies, we identify key architectural design choices such as exponential gating and bidirectionality contributing to its effectiveness. Our best xLSTM-based model, xLSTM-SENet2, outperforms state-of-the-art Mamba- and Conformer-based systems on the Voicebank+DEMAND dataset.
Abstract:In this paper, we propose several methods that incorporate vocal tract length (VTL) warped features for spoken keyword spotting (KWS). The first method, VTL-independent KWS, involves training a single deep neural network (DNN) that utilizes VTL features with various warping factors. During training, a specific VTL feature is randomly selected per epoch, allowing the exploration of VTL variations. During testing, the VTL features with different warping factors of a test utterance are scored against the DNN and combined with equal weight. In the second method scores the conventional features of a test utterance (without VTL warping) against the DNN. The third method, VTL-concatenation KWS, concatenates VTL warped features to form high-dimensional features for KWS. Evaluations carried out on the English Google Command dataset demonstrate that the proposed methods improve the accuracy of KWS.
Abstract:Target-Speaker Voice Activity Detection (TS-VAD) is the task of detecting the presence of speech from a known target-speaker in an audio frame. Recently, deep neural network-based models have shown good performance in this task. However, training these models requires extensive labelled data, which is costly and time-consuming to obtain, particularly if generalization to unseen environments is crucial. To mitigate this, we propose a causal, Self-Supervised Learning (SSL) pretraining framework, called Denoising Autoregressive Predictive Coding (DN-APC), to enhance TS-VAD performance in noisy conditions. We also explore various speaker conditioning methods and evaluate their performance under different noisy conditions. Our experiments show that DN-APC improves performance in noisy conditions, with a general improvement of approx. 2% in both seen and unseen noise. Additionally, we find that FiLM conditioning provides the best overall performance. Representation analysis via tSNE plots reveals robust initial representations of speech and non-speech from pretraining. This underscores the effectiveness of SSL pretraining in improving the robustness and performance of TS-VAD models in noisy environments.
Abstract:In this work, we present BiSSL, a first-of-its-kind training framework that introduces bilevel optimization to enhance the alignment between the pretext pre-training and downstream fine-tuning stages in self-supervised learning. BiSSL formulates the pretext and downstream task objectives as the lower- and upper-level objectives in a bilevel optimization problem and serves as an intermediate training stage within the self-supervised learning pipeline. By more explicitly modeling the interdependence of these training stages, BiSSL facilitates enhanced information sharing between them, ultimately leading to a backbone parameter initialization that is better suited for the downstream task. We propose a training algorithm that alternates between optimizing the two objectives defined in BiSSL. Using a ResNet-18 backbone pre-trained with SimCLR on the STL10 dataset, we demonstrate that our proposed framework consistently achieves improved or competitive classification accuracies across various downstream image classification datasets compared to the conventional self-supervised learning pipeline. Qualitative analyses of the backbone features further suggest that BiSSL enhances the alignment of downstream features in the backbone prior to fine-tuning.
Abstract:Automatic speech recognition (ASR) systems are known to be vulnerable to adversarial attacks. This paper addresses detection and defence against targeted white-box attacks on speech signals for ASR systems. While existing work has utilised diffusion models (DMs) to purify adversarial examples, achieving state-of-the-art results in keyword spotting tasks, their effectiveness for more complex tasks such as sentence-level ASR remains unexplored. Additionally, the impact of the number of forward diffusion steps on performance is not well understood. In this paper, we systematically investigate the use of DMs for defending against adversarial attacks on sentences and examine the effect of varying forward diffusion steps. Through comprehensive experiments on the Mozilla Common Voice dataset, we demonstrate that two forward diffusion steps can completely defend against adversarial attacks on sentences. Moreover, we introduce a novel, training-free approach for detecting adversarial attacks by leveraging a pre-trained DM. Our experimental results show that this method can detect adversarial attacks with high accuracy.
Abstract:Speech signals encompass various information across multiple levels including content, speaker, and style. Disentanglement of these information, although challenging, is important for applications such as voice conversion. The contrastive predictive coding supported factorized variational autoencoder achieves unsupervised disentanglement of a speech signal into speaker and content embeddings by assuming speaker info to be temporally more stable than content-induced variations. However, this assumption may introduce other temporal stable information into the speaker embeddings, like environment or emotion, which we call style. In this work, we propose a method to further disentangle non-content features into distinct speaker and style features, notably by leveraging readily accessible and well-defined speaker labels without the necessity for style labels. Experimental results validate the proposed method's effectiveness on extracting disentangled features, thereby facilitating speaker, style, or combined speaker-style conversion.
Abstract:While the transformer has emerged as the eminent neural architecture, several independent lines of research have emerged to address its limitations. Recurrent neural approaches have also observed a lot of renewed interest, including the extended long short-term memory (xLSTM) architecture, which reinvigorates the original LSTM architecture. However, while xLSTMs have shown competitive performance compared to the transformer, their viability for learning self-supervised general-purpose audio representations has not yet been evaluated. This work proposes Audio xLSTM (AxLSTM), an approach to learn audio representations from masked spectrogram patches in a self-supervised setting. Pretrained on the AudioSet dataset, the proposed AxLSTM models outperform comparable self-supervised audio spectrogram transformer (SSAST) baselines by up to 20% in relative performance across a set of ten diverse downstream tasks while having up to 45% fewer parameters.