Search

Recent AIGC systems possess the capability to generate digital multimedia content based on human language instructions, such as text, image and video. However, when it comes to speech, existing methods related to human instruction-to-speech generation exhibit two limitations. Firstly, they require the division of inputs into content prompt (transcript) and description prompt (style and speaker), instead of directly supporting human instruction. This division is less natural in form and does not align with other AIGC models. Secondly, the practice of utilizing an independent description prompt to model speech style, without considering the transcript content, restricts the ability to control speech at a fine-grained level. To address these limitations, we propose VoxInstruct, a novel unified multilingual codec language modeling framework that extends traditional text-to-speech tasks into a general human instruction-to-speech task. Our approach enhances the expressiveness of human instruction-guided speech generation and aligns the speech generation paradigm with other modalities. To enable the model to automatically extract the content of synthesized speech from raw text instructions, we introduce speech semantic tokens as an intermediate representation for instruction-to-content guidance. We also incorporate multiple Classifier-Free Guidance (CFG) strategies into our codec language model, which strengthens the generated speech following human instructions. Furthermore, our model architecture and training strategies allow for the simultaneous support of combining speech prompt and descriptive human instruction for expressive speech synthesis, which is a first-of-its-kind attempt. Codes, models and demos are at: https://github.com/thuhcsi/VoxInstruct., Comment: Accepted by ACM Multimedia 2024

Achieving robust speech separation for overlapping speakers in various acoustic environments with noise and reverberation remains an open challenge. Although existing datasets are available to train separators for specific scenarios, they do not effectively generalize across diverse real-world scenarios. In this paper, we present a novel data simulation pipeline that produces diverse training data from a range of acoustic environments and content, and propose new training paradigms to improve quality of a general speech separation model. Specifically, we first introduce AC-SIM, a data simulation pipeline that incorporates broad variations in both content and acoustics. Then we integrate multiple training objectives into the permutation invariant training (PIT) to enhance separation quality and generalization of the trained model. Finally, we conduct comprehensive objective and human listening experiments across separation architectures and benchmarks to validate our methods, demonstrating substantial improvement of generalization on both non-homologous and real-world test sets., Comment: In Proceedings of the 25th Annual Conference of the International Speech Communication Association, Interspeech 2024

Speech-language multi-modal learning presents a significant challenge due to the fine nuanced information inherent in speech styles. Therefore, a large-scale dataset providing elaborate comprehension of speech style is urgently needed to facilitate insightful interplay between speech audio and natural language. However, constructing such datasets presents a major trade-off between large-scale data collection and high-quality annotation. To tackle this challenge, we propose an automatic speech annotation system for expressiveness interpretation that annotates in-the-wild speech clips with expressive and vivid human language descriptions. Initially, speech audios are processed by a series of expert classifiers and captioning models to capture diverse speech characteristics, followed by a fine-tuned LLaMA for customized annotation generation. Unlike previous tag/templet-based annotation frameworks with limited information and diversity, our system provides in-depth understandings of speech style through tailored natural language descriptions, thereby enabling accurate and voluminous data generation for large model training. With this system, we create SpeechCraft, a fine-grained bilingual expressive speech dataset. It is distinguished by highly descriptive natural language style prompts, containing approximately 2,000 hours of audio data and encompassing over two million speech clips. Extensive experiments demonstrate that the proposed dataset significantly boosts speech-language task performance in stylist speech synthesis and speech style understanding., Comment: Accepted by ACM Multimedia 2024

Recent interest in Large Vision-Language Models (LVLMs) for practical applications is moderated by the significant challenge of hallucination or the inconsistency between the factual information and the generated text. In this paper, we first perform an in-depth analysis of hallucinations and discover several novel insights about how and when LVLMs hallucinate. From our analysis, we show that: (1) The community's efforts have been primarily targeted towards reducing hallucinations related to visual recognition (VR) prompts (e.g., prompts that only require describing the image), thereby ignoring hallucinations for cognitive prompts (e.g., prompts that require additional skills like reasoning on contents of the image). (2) LVLMs lack visual perception, i.e., they can see but not necessarily understand or perceive the input image. We analyze responses to cognitive prompts and show that LVLMs hallucinate due to a perception gap: although LVLMs accurately recognize visual elements in the input image and possess sufficient cognitive skills, they struggle to respond accurately and hallucinate. To overcome this shortcoming, we propose Visual Description Grounded Decoding (VDGD), a simple, robust, and training-free method for alleviating hallucinations. Specifically, we first describe the image and add it as a prefix to the instruction. Next, during auto-regressive decoding, we sample from the plausible candidates according to their KL-Divergence (KLD) to the description, where lower KLD is given higher preference. Experimental results on several benchmarks and LVLMs show that VDGD improves significantly over other baselines in reducing hallucinations. We also propose VaLLu, a benchmark for the comprehensive evaluation of the cognitive capabilities of LVLMs., Comment: Preprint. Under review. Code will be released on paper acceptance

Instruction Tuning (IT), the process of training large language models (LLMs) using instruction-response pairs, has emerged as the predominant method for transforming base pre-trained LLMs into open-domain conversational agents. While IT has achieved notable success and widespread adoption, its limitations and shortcomings remain underexplored. In this paper, through rigorous experiments and an in-depth analysis of the changes LLMs undergo through IT, we reveal various limitations of IT. In particular, we show that (1) IT fails to enhance knowledge or skills in LLMs. LoRA fine-tuning is limited to learning response initiation and style tokens, and full-parameter fine-tuning leads to knowledge degradation. (2) Copying response patterns from IT datasets derived from knowledgeable sources leads to a decline in response quality. (3) Full-parameter fine-tuning increases hallucination by inaccurately borrowing tokens from conceptually similar instances in the IT dataset for generating responses. (4) Popular methods to improve IT do not lead to performance improvements over a simple LoRA fine-tuned model. Our findings reveal that responses generated solely from pre-trained knowledge consistently outperform responses by models that learn any form of new knowledge from IT on open-source datasets. We hope the insights and challenges revealed in this paper inspire future work in related directions., Comment: Accepted to ICML 2024

This paper investigates the zero relaxation limit for general linear hyperbolic relaxation systems and establishes the asymptotic convergence of slow variables under the unimprovable weakest stability condition, akin to the Lax equivalence theorem for hyperbolic relaxation approximations. Despite potential high oscillations, the convergence of macroscopic variables is established in the strong $L^\infty_t L^2_x$ sense rather than the sense of weak convergence, time averaging, or ensemble averaging., Comment: 32 pages

A promising approach to investigating high-dimensional problems is to identify their intrinsically low-dimensional features, which can be achieved through recently developed techniques for effective low-dimensional representation of functions such as machine learning. Based on available finite-dimensional approximate solution manifolds, this paper proposes a novel model reduction framework for kinetic equations. The method employs projections onto tangent bundles of approximate manifolds, naturally resulting in first-order hyperbolic systems. Under certain conditions on the approximate manifolds, the reduced models preserve several crucial properties, including hyperbolicity, conservation laws, entropy dissipation, finite propagation speed, and linear stability. For the first time, this paper rigorously discusses the relation between the H-theorem of kinetic equations and the linear stability conditions of reduced systems, determining the choice of Riemannian metrics involved in the model reduction. The framework is widely applicable for the model reduction of many models in kinetic theory., Comment: 46 pages

Spoken language recognition (SLR) is the task of automatically identifying the language present in a speech signal. Existing SLR models are either too computationally expensive or too large to run effectively on devices with limited resources. For real-world deployment, a model should also gracefully handle unseen languages outside of the target language set, yet prior work has focused on closed-set classification where all input languages are known a-priori. In this paper we address these two limitations: we explore efficient model architectures for SLR based on convolutional networks, and propose a multilabel training strategy to handle non-target languages at inference time. Using the VoxLingua107 dataset, we show that our models obtain competitive results while being orders of magnitude smaller and faster than current state-of-the-art methods, and that our multilabel strategy is more robust to unseen non-target languages compared to multiclass classification., Comment: Accepted to InterSpeech 2023

We present new estimates of the importance of teachers in early grades for later grade outcomes, but unlike the existing literature that examines teacher "fade-out," we directly compare the contribution of early-grade teachers to later year outcomes against the contributions of later year teachers to the same later year outcomes. Where the prior literature finds that much of the contribution of early teachers fades away, we find that the contributions of early-year teachers remain important in later grades. The difference in contributions to eighth-grade outcomes between an effective and ineffective fourth-grade teacher is about half the difference among eighth-grade teachers. The effect on eighth-grade outcomes of replacing a fourth-grade teacher who is below the 5th percentile with a median teacher is about half the underrepresented minority (URM)/non-URM achievement gap. Our results reinforce earlier conclusions in the literature that teachers in all grades are important for student achievement.

Many audio processing tasks require perceptual assessment. However, the time and expense of obtaining ``gold standard'' human judgments limit the availability of such data. Most applications incorporate full reference or other similarity-based metrics (e.g. PESQ) that depend on a clean reference. Researchers have relied on such metrics to evaluate and compare various proposed methods, often concluding that small, measured differences imply one is more effective than another. This paper demonstrates several practical scenarios where similarity metrics fail to agree with human perception, because they: (1) vary with clean references; (2) rely on attributes that humans factor out when considering quality, and (3) are sensitive to imperceptible signal level differences. In those scenarios, we show that no-reference metrics do not suffer from such shortcomings and correlate better with human perception. We conclude therefore that similarity serves as an unreliable proxy for audio quality., Comment: To Appear, Interspeech 2022

States are responsible for setting and evaluating the standards that teacher preparation programs (TPPs) must meet for accreditation. Despite the considerable investment that states make in this process, no prior research has linked the ratings of TPPs generated by program reviews to inservice teacher performance. In this paper, we describe analyses of program review ratings from Massachusetts and their relationship to formal inservice teacher evaluation ratings and the value-added effectiveness of teachers. When comparisons are made across all schools and districts in the state, we find that a TPP's review scores are positively predictive of both inservice teacher evaluations and value added of TPP graduates, particularly when scores are aggregated within specific categories like partnerships and field-based practices. These relationships, however, become more modest for teacher evaluations and statistically insignificant for value added when the relationships are identified based on comparisons between TPP graduates who are teaching in the same schools and districts. It is not possible to separate whether these differences are due to the TPPs, the schools and districts themselves, or the connections between them, so future work is necessary to further validate TPP review scores in this setting and others.

Consumer-grade music recordings such as those captured by mobile devices typically contain distortions in the form of background noise, reverb, and microphone-induced EQ. This paper presents a deep learning approach to enhance low-quality music recordings by combining (i) an image-to-image translation model for manipulating audio in its mel-spectrogram representation and (ii) a music vocoding model for mapping synthetically generated mel-spectrograms to perceptually realistic waveforms. We find that this approach to music enhancement outperforms baselines which use classical methods for mel-spectrogram inversion and an end-to-end approach directly mapping noisy waveforms to clean waveforms. Additionally, in evaluating the proposed method with a listening test, we analyze the reliability of common audio enhancement evaluation metrics when used in the music domain., Comment: ICASSP 2022

What audio embedding approach generalizes best to a wide range of downstream tasks across a variety of everyday domains without fine-tuning? The aim of the HEAR benchmark is to develop a general-purpose audio representation that provides a strong basis for learning in a wide variety of tasks and scenarios. HEAR evaluates audio representations using a benchmark suite across a variety of domains, including speech, environmental sound, and music. HEAR was launched as a NeurIPS 2021 shared challenge. In the spirit of shared exchange, each participant submitted an audio embedding model following a common API that is general-purpose, open-source, and freely available to use. Twenty-nine models by thirteen external teams were evaluated on nineteen diverse downstream tasks derived from sixteen datasets. Open evaluation code, submitted models and datasets are key contributions, enabling comprehensive and reproducible evaluation, as well as previously impossible longitudinal studies. It still remains an open question whether one single general-purpose audio representation can perform as holistically as the human ear., Comment: to appear in Proceedings of Machine Learning Research (PMLR): NeurIPS 2021 Competition Track

We improve the resolvent estimate in the Kreiss matrix theorem for a set of matrices that generate uniformly bounded semigroups. The new resolvent estimate is proved to be equivalent to Kreiss's resolvent condition, and it better describes the behavior of the resolvents at infinity., Comment: 9 pages

We use teacher candidate test scores on the Massachusetts Tests for Educator Licensure (MTEL), linked to student and teacher outcomes in the state, to investigate the predictive validity of these teacher licensure tests. We find that MTEL scores are positive and statistically significant predictors of teachers' in-service performance ratings and contributions to student test scores (i.e., value added) once they enter the workforce. We then explore whether these relationships vary for candidates and teachers of color. We find that teacher candidates of color have lower first-time pass rates and are also less likely to retake licensure tests if they fail than are White teacher candidates, but we do not find consistent evidence that MTEL scores are less predictive of value added for teachers of color. Finally, we find that MTEL scores are more predictive of teacher performance ratings for teachers of color than for White teachers.

Modifying the pitch and timing of an audio signal are fundamental audio editing operations with applications in speech manipulation, audio-visual synchronization, and singing voice editing and synthesis. Thus far, methods for pitch-shifting and time-stretching that use digital signal processing (DSP) have been favored over deep learning approaches due to their speed and relatively higher quality. However, even existing DSP-based methods for pitch-shifting and time-stretching induce artifacts that degrade audio quality. In this paper, we propose Controllable LPCNet (CLPCNet), an improved LPCNet vocoder capable of pitch-shifting and time-stretching of speech. For objective evaluation, we show that CLPCNet performs pitch-shifting of speech on unseen datasets with high accuracy relative to prior neural methods. For subjective evaluation, we demonstrate that the quality and naturalness of pitch-shifting and time-stretching with CLPCNet on unseen datasets meets or exceeds competitive neural- or DSP-based approaches., Comment: Submitted to ICASSP 2022

Recent advances in deep learning have expanded possibilities to generate music, but generating a customizable full piece of music with consistent long-term structure remains a challenge. This paper introduces MusicFrameworks, a hierarchical music structure representation and a multi-step generative process to create a full-length melody guided by long-term repetitive structure, chord, melodic contour, and rhythm constraints. We first organize the full melody with section and phrase-level structure. To generate melody in each phrase, we generate rhythm and basic melody using two separate transformer-based networks, and then generate the melody conditioned on the basic melody, rhythm and chords in an auto-regressive manner. By factoring music generation into sub-problems, our approach allows simpler models and requires less data. To customize or add variety, one can alter chords, basic melody, and rhythm structure in the music frameworks, letting our networks generate the melody accordingly. Additionally, we introduce new features to encode musical positional information, rhythm patterns, and melodic contours based on musical domain knowledge. A listening test reveals that melodies generated by our method are rated as good as or better than human-composed music in the POP909 dataset about half the time., Comment: 6 pages, 9 figures, in Proc. of the 22nd Int. Society for Music Information Retrieval Conf.,Online, 2021

Text-based speech editors expedite the process of editing speech recordings by permitting editing via intuitive cut, copy, and paste operations on a speech transcript. A major drawback of current systems, however, is that edited recordings often sound unnatural because of prosody mismatches around edited regions. In our work, we propose a new context-aware method for more natural sounding text-based editing of speech. To do so, we 1) use a series of neural networks to generate salient prosody features that are dependent on the prosody of speech surrounding the edit and amenable to fine-grained user control 2) use the generated features to control a standard pitch-shift and time-stretch method and 3) apply a denoising neural network to remove artifacts induced by the signal manipulation to yield a high-fidelity result. We evaluate our approach using a subjective listening test, provide a detailed comparative analysis, and conclude several interesting insights., Comment: To appear in proceedings of ICASSP 2021

Many speech processing methods based on deep learning require an automatic and differentiable audio metric for the loss function. The DPAM approach of Manocha et al. learns a full-reference metric trained directly on human judgments, and thus correlates well with human perception. However, it requires a large number of human annotations and does not generalize well outside the range of perturbations on which it was trained. This paper introduces CDPAM, a metric that builds on and advances DPAM. The primary improvement is to combine contrastive learning and multi-dimensional representations to build robust models from limited data. In addition, we collect human judgments on triplet comparisons to improve generalization to a broader range of audio perturbations. CDPAM correlates well with human responses across nine varied datasets. We also show that adding this metric to existing speech synthesis and enhancement methods yields significant improvement, as measured by objective and subjective tests., Comment: Dataset, code and sound examples can be found at https://github.com/pranaymanocha/PerceptualAudio/tree/master/cdpam

We propose a high order numerical homogenization method for dissipative ordinary differential equations (ODEs) containing two time scales. Essentially, only first order homogenized model globally in time can be derived. To achieve a high order method, we have to adopt a numerical approach in the framework of the heterogeneous multiscale method (HMM). By a successively refined microscopic solver, the accuracy improvement up to arbitrary order is attained providing input data smooth enough. Based on the formulation of the high order microscopic solver we derived, an iterative formula to calculate the microscopic solver is then proposed. Using the iterative formula, we develop an implementation to the method in an efficient way for practical applications. Several numerical examples are presented to validate the new models and numerical methods., Comment: 29 pages, 8 figures

Music similarity search is useful for a variety of creative tasks such as replacing one music recording with another recording with a similar "feel", a common task in video editing. For this task, it is typically necessary to define a similarity metric to compare one recording to another. Music similarity, however, is hard to define and depends on multiple simultaneous notions of similarity (i.e. genre, mood, instrument, tempo). While prior work ignore this issue, we embrace this idea and introduce the concept of multidimensional similarity and unify both global and specialized similarity metrics into a single, semantically disentangled multidimensional similarity metric. To do so, we adapt a variant of deep metric learning called conditional similarity networks to the audio domain and extend it using track-based information to control the specificity of our model. We evaluate our method and show that our single, multidimensional model outperforms both specialized similarity spaces and alternative baselines. We also run a user-study and show that our approach is favored by human annotators as well., Comment: Accepted for publication at the 45th International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2020)

Deep representation learning offers a powerful paradigm for mapping input data onto an organized embedding space and is useful for many music information retrieval tasks. Two central methods for representation learning include deep metric learning and classification, both having the same goal of learning a representation that can generalize well across tasks. Along with generalization, the emerging concept of disentangled representations is also of great interest, where multiple semantic concepts (e.g., genre, mood, instrumentation) are learned jointly but remain separable in the learned representation space. In this paper we present a single representation learning framework that elucidates the relationship between metric learning, classification, and disentanglement in a holistic manner. For this, we (1) outline past work on the relationship between metric learning and classification, (2) extend this relationship to multi-label data by exploring three different learning approaches and their disentangled versions, and (3) evaluate all models on four tasks (training time, similarity retrieval, auto-tagging, and triplet prediction). We find that classification-based models are generally advantageous for training time, similarity retrieval, and auto-tagging, while deep metric learning exhibits better performance for triplet-prediction. Finally, we show that our proposed approach yields state-of-the-art results for music auto-tagging., Comment: Accepted for publication at the 21st International Society for Music Information Retrieval Conference (ISMIR 2020)

Speech synthesis has recently seen significant improvements in fidelity, driven by the advent of neural vocoders and neural prosody generators. However, these systems lack intuitive user controls over prosody, making them unable to rectify prosody errors (e.g., misplaced emphases and contextually inappropriate emotions) or generate prosodies with diverse speaker excitement levels and emotions. We address these limitations with a user-controllable, context-aware neural prosody generator. Given a real or synthesized speech recording, our model allows a user to input prosody constraints for certain time frames and generates the remaining time frames from input text and contextual prosody. We also propose a pitch-shifting neural vocoder to modify input speech to match the synthesized prosody. Through objective and subjective evaluations we show that we can successfully incorporate user control into our prosody generation model without sacrificing the overall naturalness of the synthesized speech., Comment: To appear in proceedings of INTERSPEECH 2020

Many reinforcement learning methods achieve great success in practice but lack theoretical foundation. In this paper, we study the convergence analysis on the problem of the Linear Quadratic Regulator (LQR). The global linear convergence properties and sample complexities are established for several popular algorithms such as the policy gradient algorithm, TD-learning and the actor-critic (AC) algorithm. Our results show that the actor-critic algorithm can reduce the sample complexity compared with the policy gradient algorithm. Although our analysis is still preliminary, it explains the benefit of AC algorithm in a certain sense.

Real-world audio recordings are often degraded by factors such as noise, reverberation, and equalization distortion. This paper introduces HiFi-GAN, a deep learning method to transform recorded speech to sound as though it had been recorded in a studio. We use an end-to-end feed-forward WaveNet architecture, trained with multi-scale adversarial discriminators in both the time domain and the time-frequency domain. It relies on the deep feature matching losses of the discriminators to improve the perceptual quality of enhanced speech. The proposed model generalizes well to new speakers, new speech content, and new environments. It significantly outperforms state-of-the-art baseline methods in both objective and subjective experiments., Comment: Accepted by INTERSPEECH 2020

Non-parallel many-to-many voice conversion remains an interesting but challenging speech processing task. Many style-transfer-inspired methods such as generative adversarial networks (GANs) and variational autoencoders (VAEs) have been proposed. Recently, AutoVC, a conditional autoencoders (CAEs) based method achieved state-of-the-art results by disentangling the speaker identity and speech content using information-constraining bottlenecks, and it achieves zero-shot conversion by swapping in a different speaker's identity embedding to synthesize a new voice. However, we found that while speaker identity is disentangled from speech content, a significant amount of prosodic information, such as source F0, leaks through the bottleneck, causing target F0 to fluctuate unnaturally. Furthermore, AutoVC has no control of the converted F0 and thus unsuitable for many applications. In the paper, we modified and improved autoencoder-based voice conversion to disentangle content, F0, and speaker identity at the same time. Therefore, we can control the F0 contour, generate speech with F0 consistent with the target speaker, and significantly improve quality and similarity. We support our improvement through quantitative and qualitative analysis.

Many audio processing tasks require perceptual assessment. The ``gold standard`` of obtaining human judgments is time-consuming, expensive, and cannot be used as an optimization criterion. On the other hand, automated metrics are efficient to compute but often correlate poorly with human judgment, particularly for audio differences at the threshold of human detection. In this work, we construct a metric by fitting a deep neural network to a new large dataset of crowdsourced human judgments. Subjects are prompted to answer a straightforward, objective question: are two recordings identical or not? These pairs are algorithmically generated under a variety of perturbations, including noise, reverb, and compression artifacts; the perturbation space is probed with the goal of efficiently identifying the just-noticeable difference (JND) level of the subject. We show that the resulting learned metric is well-calibrated with human judgments, outperforming baseline methods. Since it is a deep network, the metric is differentiable, making it suitable as a loss function for other tasks. Thus, simply replacing an existing loss (e.g., deep feature loss) with our metric yields significant improvement in a denoising network, as measured by subjective pairwise comparison., Comment: Dataset, code and sound examples can be found at https://pixl.cs.princeton.edu/pubs/Manocha_2020_ADP/

Editing talking-head video to change the speech content or to remove filler words is challenging. We propose a novel method to edit talking-head video based on its transcript to produce a realistic output video in which the dialogue of the speaker has been modified, while maintaining a seamless audio-visual flow (i.e. no jump cuts). Our method automatically annotates an input talking-head video with phonemes, visemes, 3D face pose and geometry, reflectance, expression and scene illumination per frame. To edit a video, the user has to only edit the transcript, and an optimization strategy then chooses segments of the input corpus as base material. The annotated parameters corresponding to the selected segments are seamlessly stitched together and used to produce an intermediate video representation in which the lower half of the face is rendered with a parametric face model. Finally, a recurrent video generation network transforms this representation to a photorealistic video that matches the edited transcript. We demonstrate a large variety of edits, such as the addition, removal, and alteration of words, as well as convincing language translation and full sentence synthesis., Comment: A version with higher resolution images can be downloaded from the authors' website