Hao Tang

Hui Wei, Zhixiang Wang, Xuemei Jia et al.

Adversarial attacks on thermal infrared imaging expose the risk of related applications. Estimating the security of these systems is essential for safely deploying them in the real world. In many cases, realizing the attacks in the physical space requires elaborate special perturbations. These solutions are often \emph{impractical} and \emph{attention-grabbing}. To address the need for a physically practical and stealthy adversarial attack, we introduce \textsc{HotCold} Block, a novel physical attack for infrared detectors that hide persons utilizing the wearable Warming Paste and Cooling Paste. By attaching these readily available temperature-controlled materials to the body, \textsc{HotCold} Block evades human eyes efficiently. Moreover, unlike existing methods that build adversarial patches with complex texture and structure features, \textsc{HotCold} Block utilizes an SSP-oriented adversarial optimization algorithm that enables attacks with pure color blocks and explores the influence of size, shape, and position on attack performance. Extensive experimental results in both digital and physical environments demonstrate the performance of our proposed \textsc{HotCold} Block. \emph{Code is available: \textcolor{magenta}{https://github.com/weihui1308/HOTCOLDBlock}}.

Tzu-Quan Lin, Hung-yi Lee, Hao Tang

Self-supervised models have had great success in learning speech representations that can generalize to various downstream tasks. However, most self-supervised models require a large amount of compute and multiple GPUs to train, significantly hampering the development of self-supervised learning. In an attempt to reduce the computation of training, we revisit the training of HuBERT, a highly successful self-supervised model. We improve and simplify several key components, including the loss function, input representation, and training in multiple stages. Our model, MelHuBERT, is able to achieve favorable performance on phone recognition, speaker identification, and automatic speech recognition against HuBERT, while saving 31.2% of the pre-training time, or equivalently 33.5% MACs per one second speech. The code and pre-trained models are available in https://github.com/nervjack2/MelHuBERT.

Yen Meng, Hsuan-Jui Chen, Jiatong Shi et al.

Compressing self-supervised models has become increasingly necessary, as self-supervised models become larger. While previous approaches have primarily focused on compressing the model size, shortening sequences is also effective in reducing the computational cost. In this work, we study fixed-length and variable-length subsampling along the time axis in self-supervised learning. We explore how individual downstream tasks are sensitive to input frame rates. Subsampling while training self-supervised models not only improves the overall performance on downstream tasks under certain frame rates, but also brings significant speed-up in inference. Variable-length subsampling performs particularly well under low frame rates. In addition, if we have access to phonetic boundaries, we find no degradation in performance for an average frame rate as low as 10 Hz.

Ramon Sanabria, Hao Tang, Sharon Goldwater

Given the strong results of self-supervised models on various tasks, there have been surprisingly few studies exploring self-supervised representations for acoustic word embeddings (AWE), fixed-dimensional vectors representing variable-length spoken word segments. In this work, we study several pre-trained models and pooling methods for constructing AWEs with self-supervised representations. Owing to the contextualized nature of self-supervised representations, we hypothesize that simple pooling methods, such as averaging, might already be useful for constructing AWEs. When evaluating on a standard word discrimination task, we find that HuBERT representations with mean-pooling rival the state of the art on English AWEs. More surprisingly, despite being trained only on English, HuBERT representations evaluated on Xitsonga, Mandarin, and French consistently outperform the multilingual model XLSR-53 (as well as Wav2Vec 2.0 trained on English).

Tzu-Quan Lin, Tsung-Huan Yang, Chun-Yao Chang et al.

Transformer-based self-supervised models have achieved remarkable success in speech processing, but their large size and high inference cost present significant challenges for real-world deployment. While numerous compression techniques have been proposed, inconsistent evaluation metrics make it difficult to compare their practical effectiveness. In this work, we conduct a comprehensive study of four common compression methods, including weight pruning, head pruning, low-rank approximation, and knowledge distillation on self-supervised speech Transformers. We evaluate each method under three key metrics: parameter count, multiply-accumulate operations, and real-time factor. Results show that each method offers distinct advantages. In addition, we contextualize recent compression techniques, comparing DistilHuBERT, FitHuBERT, LightHuBERT, ARMHuBERT, and STaRHuBERT under the same framework, offering practical guidance on compression for deployment.

Ramon Sanabria, Ondrej Klejch, Hao Tang et al.

Acoustic word embeddings are typically created by training a pooling function using pairs of word-like units. For unsupervised systems, these are mined using k-nearest neighbor (KNN) search, which is slow. Recently, mean-pooled representations from a pre-trained self-supervised English model were suggested as a promising alternative, but their performance on target languages was not fully competitive. Here, we explore improvements to both approaches: we use continued pre-training to adapt the self-supervised model to the target language, and we use a multilingual phone recognizer (MPR) to mine phone n-gram pairs for training the pooling function. Evaluating on four languages, we show that both methods outperform a recent approach on word discrimination. Moreover, the MPR method is orders of magnitude faster than KNN, and is highly data efficient. We also show a small improvement from performing learned pooling on top of the continued pre-trained representations.

Gene-Ping Yang, Hao Tang

Attention mechanism in sequence-to-sequence models is designed to model the alignments between acoustic features and output tokens in speech recognition. However, attention weights produced by models trained end to end do not always correspond well with actual alignments, and several studies have further argued that attention weights might not even correspond well with the relevance attribution of frames. Regardless, visual similarity between attention weights and alignments is widely used during training as an indicator of the models quality. In this paper, we treat the correspondence between attention weights and alignments as a learning problem by imposing a supervised attention loss. Experiments have shown significant improved performance, suggesting that learning the alignments well during training critically determines the performance of sequence-to-sequence models.

Sung-Lin Yeh, Hao Tang

While discrete latent variable models have had great success in self-supervised learning, most models assume that frames are independent. Due to the segmental nature of phonemes in speech perception, modeling dependencies among latent variables at the frame level can potentially improve the learned representations on phonetic-related tasks. In this work, we assume Markovian dependencies among latent variables, and propose to learn speech representations with neural hidden Markov models. Our general framework allows us to compare to self-supervised models that assume independence, while keeping the number of parameters fixed. The added dependencies improve the accessibility of phonetic information, phonetic segmentation, and the cluster purity of phones, showcasing the benefit of the assumed dependencies.

Jianing Qi, Jiawei Liu, Hao Tang et al.

Vision Language Models (VLMs) excel at identifying and describing objects but often fail at spatial reasoning. We study why VLMs, such as LLaVA, underutilize spatial cues despite having positional encodings and spatially rich vision encoder features. Our analysis reveals a key imbalance: vision token embeddings have much larger norms than text tokens, suppressing LLM's position embedding. To expose this mechanism, we developed three interpretability tools: (1) the Position Sensitivity Index, which quantifies reliance on token order, (2) the Cross Modality Balance, which reveals attention head allocation patterns, and (3) a RoPE Sensitivity probe, which measures dependence on rotary positional embeddings. These tools uncover that vision tokens and system prompts dominate attention. We validated our mechanistic understanding through targeted interventions that predictably restore positional sensitivity. These findings reveal previously unknown failure modes in multimodal attention and demonstrate how interpretability analysis can guide principled improvements.

Tzu-Quan Lin, Hung-yi Lee, Hao Tang

Self-supervised speech models have shown to be useful for various tasks, but their large size limits the use in devices with low computing power and memory. In this work, we explore early exit, an approach for reducing latency by exiting the forward process of a network early. Most approaches of early exit need a separate early exit model for each task, with some even requiring fine-tuning of the entire pretrained model. We introduce Data Adaptive Self-Supervised Early Exit (DAISY), an approach that decides when to exit based on the self-supervised loss, eliminating the need for multiple round of training and fine-tuning. DAISY matches the performance of HuBERT on the MiniSUPERB benchmark, but with much faster inference times. Our analysis on the adaptivity of DAISY shows that the model exits early (using fewer layers) on clean data while exits late (using more layers) on noisy data, dynamically adjusting the computational cost of inference based on the noise level of each sample.

Sung-Lin Yeh, Hao Tang

While several self-supervised approaches for learning discrete speech representation have been proposed, it is unclear how these seemingly similar approaches relate to each other. In this paper, we consider a generative model with discrete latent variables that learns a discrete representation for speech. The objective of learning the generative model is formulated as information-theoretic co-training. Besides the wide generality, the objective can be optimized with several approaches, subsuming HuBERT-like training and vector quantization for learning discrete representation. Empirically, we find that the proposed approach learns discrete representation that is highly correlated with phonetic units, more correlated than HuBERT-like training and vector quantization.

Yu-An Chung, Hao Tang, James Glass

Autoregressive Predictive Coding (APC), as a self-supervised objective, has enjoyed success in learning representations from large amounts of unlabeled data, and the learned representations are rich for many downstream tasks. However, the connection between low self-supervised loss and strong performance in downstream tasks remains unclear. In this work, we propose Vector-Quantized Autoregressive Predictive Coding (VQ-APC), a novel model that produces quantized representations, allowing us to explicitly control the amount of information encoded in the representations. By studying a sequence of increasingly limited models, we reveal the constituents of the learned representations. In particular, we confirm the presence of information with probing tasks, while showing the absence of information with mutual information, uncovering the model's preference in preserving speech information as its capacity becomes constrained. We find that there exists a point where phonetic and speaker information are amplified to maximize a self-supervised objective. As a byproduct, the learned codes for a particular model capacity correspond well to English phones.

Yu-An Chung, Wei-Ning Hsu, Hao Tang et al.

This paper proposes a novel unsupervised autoregressive neural model for learning generic speech representations. In contrast to other speech representation learning methods that aim to remove noise or speaker variabilities, ours is designed to preserve information for a wide range of downstream tasks. In addition, the proposed model does not require any phonetic or word boundary labels, allowing the model to benefit from large quantities of unlabeled data. Speech representations learned by our model significantly improve performance on both phone classification and speaker verification over the surface features and other supervised and unsupervised approaches. Further analysis shows that different levels of speech information are captured by our model at different layers. In particular, the lower layers tend to be more discriminative for speakers, while the upper layers provide more phonetic content.

Hao Tang, James Glass

Acoustics-to-word models are end-to-end speech recognizers that use words as targets without relying on pronunciation dictionaries or graphemes. These models are notoriously difficult to train due to the lack of linguistic knowledge. It is also unclear how the amount of training data impacts the optimization and generalization of such models. In this work, we study the optimization and generalization of acoustics-to-word models under different amounts of training data. In addition, we study three types of inductive bias, leveraging a pronunciation dictionary, word boundary annotations, and constraints on word durations. We find that constraining word durations leads to the most improvement. Finally, we analyze the word embedding space learned by the model, and find that the space has a structure dominated by the pronunciation of words. This suggests that the contexts of words, instead of their phonetic structure, should be the future focus of inductive bias in acoustics-to-word models.

Suwon Shon, Hao Tang, James Glass

In this paper, we propose a Convolutional Neural Network (CNN) based speaker recognition model for extracting robust speaker embeddings. The embedding can be extracted efficiently with linear activation in the embedding layer. To understand how the speaker recognition model operates with text-independent input, we modify the structure to extract frame-level speaker embeddings from each hidden layer. We feed utterances from the TIMIT dataset to the trained network and use several proxy tasks to study the networks ability to represent speech input and differentiate voice identity. We found that the networks are better at discriminating broad phonetic classes than individual phonemes. In particular, frame-level embeddings that belong to the same phonetic classes are similar (based on cosine distance) for the same speaker. The frame level representation also allows us to analyze the networks at the frame level, and has the potential for other analyses to improve speaker recognition.

Hao Tang, James Glass

Recurrent neural networks have been the dominant models for many speech and language processing tasks. However, we understand little about the behavior and the class of functions recurrent networks can realize. Moreover, the heuristics used during training complicate the analyses. In this paper, we study recurrent networks' ability to learn long-term dependency in the context of speech recognition. We consider two decoding approaches, online and batch decoding, and show the classes of functions to which the decoding approaches correspond. We then draw a connection between batch decoding and a popular training approach for recurrent networks, truncated backpropagation through time. Changing the decoding approach restricts the amount of past history recurrent networks can use for prediction, allowing us to analyze their ability to remember. Empirically, we utilize long-term dependency in subphonetic states, phonemes, and words, and show how the design decisions, such as the decoding approach, lookahead, context frames, and consecutive prediction, characterize the behavior of recurrent networks. Finally, we draw a connection between Markov processes and vanishing gradients. These results have implications for studying the long-term dependency in speech data and how these properties are learned by recurrent networks.

Wei-Ning Hsu, Hao Tang, James Glass

The current trend in automatic speech recognition is to leverage large amounts of labeled data to train supervised neural network models. Unfortunately, obtaining data for a wide range of domains to train robust models can be costly. However, it is relatively inexpensive to collect large amounts of unlabeled data from domains that we want the models to generalize to. In this paper, we propose a novel unsupervised adaptation method that learns to synthesize labeled data for the target domain from unlabeled in-domain data and labeled out-of-domain data. We first learn without supervision an interpretable latent representation of speech that encodes linguistic and nuisance factors (e.g., speaker and channel) using different latent variables. To transform a labeled out-of-domain utterance without altering its transcript, we transform the latent nuisance variables while maintaining the linguistic variables. To demonstrate our approach, we focus on a channel mismatch setting, where the domain of interest is distant conversational speech, and labels are only available for close-talking speech. Our proposed method is evaluated on the AMI dataset, outperforming all baselines and bridging the gap between unadapted and in-domain models by over 77% without using any parallel data.

Hao Tang, Wei-Ning Hsu, Francois Grondin et al.

Speech recognizers trained on close-talking speech do not generalize to distant speech and the word error rate degradation can be as large as 40% absolute. Most studies focus on tackling distant speech recognition as a separate problem, leaving little effort to adapting close-talking speech recognizers to distant speech. In this work, we review several approaches from a domain adaptation perspective. These approaches, including speech enhancement, multi-condition training, data augmentation, and autoencoders, all involve a transformation of the data between domains. We conduct experiments on the AMI data set, where these approaches can be realized under the same controlled setting. These approaches lead to different amounts of improvement under their respective assumptions. The purpose of this paper is to quantify and characterize the performance gap between the two domains, setting up the basis for studying adaptation of speech recognizers from close-talking speech to distant speech. Our results also have implications for improving distant speech recognition.

Hao Tang

Segments that span contiguous parts of inputs, such as phonemes in speech, named entities in sentences, actions in videos, occur frequently in sequence prediction problems. Segmental models, a class of models that explicitly hypothesizes segments, have allowed the exploration of rich segment features for sequence prediction. However, segmental models suffer from slow decoding, hampering the use of computationally expensive features. In this thesis, we introduce discriminative segmental cascades, a multi-pass inference framework that allows us to improve accuracy by adding higher-order features and neural segmental features while maintaining efficiency. We also show that instead of including more features to obtain better accuracy, segmental cascades can be used to speed up training and decoding. Segmental models, similarly to conventional speech recognizers, are typically trained in multiple stages. In the first stage, a frame classifier is trained with manual alignments, and then in the second stage, segmental models are trained with manual alignments and the out- puts of the frame classifier. However, obtaining manual alignments are time-consuming and expensive. We explore end-to-end training for segmental models with various loss functions, and show how end-to-end training with marginal log loss can eliminate the need for detailed manual alignments. We draw the connections between the marginal log loss and a popular end-to-end training approach called connectionist temporal classification. We present a unifying framework for various end-to-end graph search-based models, such as hidden Markov models, connectionist temporal classification, and segmental models. Finally, we discuss possible extensions of segmental models to large-vocabulary sequence prediction tasks.

Hao Tang, Liang Lu, Lingpeng Kong et al.

Segmental models are an alternative to frame-based models for sequence prediction, where hypothesized path weights are based on entire segment scores rather than a single frame at a time. Neural segmental models are segmental models that use neural network-based weight functions. Neural segmental models have achieved competitive results for speech recognition, and their end-to-end training has been explored in several studies. In this work, we review neural segmental models, which can be viewed as consisting of a neural network-based acoustic encoder and a finite-state transducer decoder. We study end-to-end segmental models with different weight functions, including ones based on frame-level neural classifiers and on segmental recurrent neural networks. We study how reducing the search space size impacts performance under different weight functions. We also compare several loss functions for end-to-end training. Finally, we explore training approaches, including multi-stage vs. end-to-end training and multitask training that combines segmental and frame-level losses.

Shubham Toshniwal, Hao Tang, Liang Lu et al.

End-to-end training of deep learning-based models allows for implicit learning of intermediate representations based on the final task loss. However, the end-to-end approach ignores the useful domain knowledge encoded in explicit intermediate-level supervision. We hypothesize that using intermediate representations as auxiliary supervision at lower levels of deep networks may be a good way of combining the advantages of end-to-end training and more traditional pipeline approaches. We present experiments on conversational speech recognition where we use lower-level tasks, such as phoneme recognition, in a multitask training approach with an encoder-decoder model for direct character transcription. We compare multiple types of lower-level tasks and analyze the effects of the auxiliary tasks. Our results on the Switchboard corpus show that this approach improves recognition accuracy over a standard encoder-decoder model on the Eval2000 test set.

Hao Tang, Weiran Wang, Kevin Gimpel et al.

Recent work on discriminative segmental models has shown that they can achieve competitive speech recognition performance, using features based on deep neural frame classifiers. However, segmental models can be more challenging to train than standard frame-based approaches. While some segmental models have been successfully trained end to end, there is a lack of understanding of their training under different settings and with different losses. We investigate a model class based on recent successful approaches, consisting of a linear model that combines segmental features based on an LSTM frame classifier. Similarly to hybrid HMM-neural network models, segmental models of this class can be trained in two stages (frame classifier training followed by linear segmental model weight training), end to end (joint training of both frame classifier and linear weights), or with end-to-end fine-tuning after two-stage training. We study segmental models trained end to end with hinge loss, log loss, latent hinge loss, and marginal log loss. We consider several losses for the case where training alignments are available as well as where they are not. We find that in general, marginal log loss provides the most consistent strong performance without requiring ground-truth alignments. We also find that training with dropout is very important in obtaining good performance with end-to-end training. Finally, the best results are typically obtained by a combination of two-stage training and fine-tuning.

Taehwan Kim, Jonathan Keane, Weiran Wang et al.

We study the problem of recognizing video sequences of fingerspelled letters in American Sign Language (ASL). Fingerspelling comprises a significant but relatively understudied part of ASL. Recognizing fingerspelling is challenging for a number of reasons: It involves quick, small motions that are often highly coarticulated; it exhibits significant variation between signers; and there has been a dearth of continuous fingerspelling data collected. In this work we collect and annotate a new data set of continuous fingerspelling videos, compare several types of recognizers, and explore the problem of signer variation. Our best-performing models are segmental (semi-Markov) conditional random fields using deep neural network-based features. In the signer-dependent setting, our recognizers achieve up to about 92% letter accuracy. The multi-signer setting is much more challenging, but with neural network adaptation we achieve up to 83% letter accuracies in this setting.

Hao Tang, Weiran Wang, Kevin Gimpel et al.

Discriminative segmental models offer a way to incorporate flexible feature functions into speech recognition. However, their appeal has been limited by their computational requirements, due to the large number of possible segments to consider. Multi-pass cascades of segmental models introduce features of increasing complexity in different passes, where in each pass a segmental model rescores lattices produced by a previous (simpler) segmental model. In this paper, we explore several ways of making segmental cascades efficient and practical: reducing the feature set in the first pass, frame subsampling, and various pruning approaches. In experiments on phonetic recognition, we find that with a combination of such techniques, it is possible to maintain competitive performance while greatly reducing decoding, pruning, and training time.

Taehwan Kim, Weiran Wang, Hao Tang et al.

We study the problem of recognition of fingerspelled letter sequences in American Sign Language in a signer-independent setting. Fingerspelled sequences are both challenging and important to recognize, as they are used for many content words such as proper nouns and technical terms. Previous work has shown that it is possible to achieve almost 90% accuracies on fingerspelling recognition in a signer-dependent setting. However, the more realistic signer-independent setting presents challenges due to significant variations among signers, coupled with the dearth of available training data. We investigate this problem with approaches inspired by automatic speech recognition. We start with the best-performing approaches from prior work, based on tandem models and segmental conditional random fields (SCRFs), with features based on deep neural network (DNN) classifiers of letters and phonological features. Using DNN adaptation, we find that it is possible to bridge a large part of the gap between signer-dependent and signer-independent performance. Using only about 115 transcribed words for adaptation from the target signer, we obtain letter accuracies of up to 82.7% with frame-level adaptation labels and 69.7% with only word labels.

Hao Tang, Weiran Wang, Kevin Gimpel et al.

Discriminative segmental models, such as segmental conditional random fields (SCRFs) and segmental structured support vector machines (SSVMs), have had success in speech recognition via both lattice rescoring and first-pass decoding. However, such models suffer from slow decoding, hampering the use of computationally expensive features, such as segment neural networks or other high-order features. A typical solution is to use approximate decoding, either by beam pruning in a single pass or by beam pruning to generate a lattice followed by a second pass. In this work, we study discriminative segmental models trained with a hinge loss (i.e., segmental structured SVMs). We show that beam search is not suitable for learning rescoring models in this approach, though it gives good approximate decoding performance when the model is already well-trained. Instead, we consider an approach inspired by structured prediction cascades, which use max-marginal pruning to generate lattices. We obtain a high-accuracy phonetic recognition system with several expensive feature types: a segment neural network, a second-order language model, and second-order phone boundary features.