Search

Various visual foundation models have distinct strengths and weaknesses, both of which can be improved through heterogeneous multi-teacher knowledge distillation without labels, termed "agglomerative models." We build upon this body of work by studying the effect of the teachers' activation statistics, particularly the impact of the loss function on the resulting student model quality. We explore a standard toolkit of statistical normalization techniques to better align the different distributions and assess their effects. Further, we examine the impact on downstream teacher-matching metrics, which motivates the use of Hadamard matrices. With these matrices, we demonstrate useful properties, showing how they can be used for isotropic standardization, where each dimension of a multivariate distribution is standardized using the same scale. We call this technique "PHI Standardization" (PHI-S) and empirically demonstrate that it produces the best student model across the suite of methods studied.

We introduce NVLM 1.0, a family of frontier-class multimodal large language models (LLMs) that achieve state-of-the-art results on vision-language tasks, rivaling the leading proprietary models (e.g., GPT-4o) and open-access models (e.g., Llama 3-V 405B and InternVL 2). Remarkably, NVLM 1.0 shows improved text-only performance over its LLM backbone after multimodal training. In terms of model design, we perform a comprehensive comparison between decoder-only multimodal LLMs (e.g., LLaVA) and cross-attention-based models (e.g., Flamingo). Based on the strengths and weaknesses of both approaches, we propose a novel architecture that enhances both training efficiency and multimodal reasoning capabilities. Furthermore, we introduce a 1-D tile-tagging design for tile-based dynamic high-resolution images, which significantly boosts performance on multimodal reasoning and OCR-related tasks. Regarding training data, we meticulously curate and provide detailed information on our multimodal pretraining and supervised fine-tuning datasets. Our findings indicate that dataset quality and task diversity are more important than scale, even during the pretraining phase, across all architectures. Notably, we develop production-grade multimodality for the NVLM-1.0 models, enabling them to excel in vision-language tasks while maintaining and even improving text-only performance compared to their LLM backbones. To achieve this, we craft and integrate a high-quality text-only dataset into multimodal training, alongside a substantial amount of multimodal math and reasoning data, leading to enhanced math and coding capabilities across modalities. To advance research in the field, we release the model weights at https://huggingface.co/nvidia/NVLM-D-72B and will open-source the training code for the community soon., Comment: Fixed the typos. For more information, please visit our project page at: https://research.nvidia.com/labs/adlr/NVLM-1

It is well established that listening to music is an issue for those with hearing loss, and hearing aids are not a universal solution. How can machine learning be used to address this? This paper details the first application of the open challenge methodology to use machine learning to improve audio quality of music for those with hearing loss. The first challenge was a stand-alone competition (CAD1) and had 9 entrants. The second was an 2024 ICASSP grand challenge (ICASSP24) and attracted 17 entrants. The challenge tasks concerned demixing and remixing pop/rock music to allow a personalised rebalancing of the instruments in the mix, along with amplification to correct for raised hearing thresholds. The software baselines provided for entrants to build upon used two state-of-the-art demix algorithms: Hybrid Demucs and Open-Unmix. Evaluation of systems was done using the objective metric HAAQI, the Hearing-Aid Audio Quality Index. No entrants improved on the best baseline in CAD1 because there was insufficient room for improvement. Consequently, for ICASSP24 the scenario was made more difficult by using loudspeaker reproduction and specified gains to be applied before remixing. This also made the scenario more useful for listening through hearing aids. 9 entrants scored better than the the best ICASSP24 baseline. Most entrants used a refined version of Hybrid Demucs and NAL-R amplification. The highest scoring system combined the outputs of several demixing algorithms in an ensemble approach. These challenges are now open benchmarks for future research with the software and data being freely available.

Machine learning techniques are an active area of research for speech enhancement for hearing aids, with one particular focus on improving the intelligibility of a noisy speech signal. Recent work has shown that feature encodings from self-supervised speech representation models can effectively capture speech intelligibility. In this work, it is shown that the distance between self-supervised speech representations of clean and noisy speech correlates more strongly with human intelligibility ratings than other signal-based metrics. Experiments show that training a speech enhancement model using this distance as part of a loss function improves the performance over using an SNR-based loss function, demonstrated by an increase in HASPI, STOI, PESQ and SI-SNR scores. This method takes inference of a high parameter count model only at training time, meaning the speech enhancement model can remain smaller, as is required for hearing aids., Comment: Accepted for EUSIPCO 2024

Supervised models for speech enhancement are trained using artificially generated mixtures of clean speech and noise signals. However, the synthetic training conditions may not accurately reflect real-world conditions encountered during testing. This discrepancy can result in poor performance when the test domain significantly differs from the synthetic training domain. To tackle this issue, the UDASE task of the 7th CHiME challenge aimed to leverage real-world noisy speech recordings from the test domain for unsupervised domain adaptation of speech enhancement models. Specifically, this test domain corresponds to the CHiME-5 dataset, characterized by real multi-speaker and conversational speech recordings made in noisy and reverberant domestic environments, for which ground-truth clean speech signals are not available. In this paper, we present the objective and subjective evaluations of the systems that were submitted to the CHiME-7 UDASE task, and we provide an analysis of the results. This analysis reveals a limited correlation between subjective ratings and several supervised nonintrusive performance metrics recently proposed for speech enhancement. Conversely, the results suggest that more traditional intrusive objective metrics can be used for in-domain performance evaluation using the reverberant LibriCHiME-5 dataset developed for the challenge. The subjective evaluation indicates that all systems successfully reduced the background noise, but always at the expense of increased distortion. Out of the four speech enhancement methods evaluated subjectively, only one demonstrated an improvement in overall quality compared to the unprocessed noisy speech, highlighting the difficulty of the task. The tools and audio material created for the CHiME-7 UDASE task are shared with the community.

Neural networks have been successfully used for non-intrusive speech intelligibility prediction. Recently, the use of feature representations sourced from intermediate layers of pre-trained self-supervised and weakly-supervised models has been found to be particularly useful for this task. This work combines the use of Whisper ASR decoder layer representations as neural network input features with an exemplar-based, psychologically motivated model of human memory to predict human intelligibility ratings for hearing-aid users. Substantial performance improvement over an established intrusive HASPI baseline system is found, including on enhancement systems and listeners unseen in the training data, with a root mean squared error of 25.3 compared with the baseline of 28.7., Comment: Accepted paper. IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), Seoul, Korea, April 2024

This paper reports on the design and outcomes of the ICASSP SP Clarity Challenge: Speech Enhancement for Hearing Aids. The scenario was a listener attending to a target speaker in a noisy, domestic environment. There were multiple interferers and head rotation by the listener. The challenge extended the second Clarity Enhancement Challenge (CEC2) by fixing the amplification stage of the hearing aid; evaluating with a combined metric for speech intelligibility and quality; and providing two evaluation sets, one based on simulation and the other on real-room measurements. Five teams improved on the baseline system for the simulated evaluation set, but the performance on the measured evaluation set was much poorer. Investigations are on-going to determine the exact cause of the mismatch between the simulated and measured data sets. The presence of transducer noise in the measurements, lower order Ambisonics harming the ability for systems to exploit binaural cues and the differences between real and simulated room impulse responses are suggested causes, Comment: ICASSP 2023

This paper describes two intelligibility prediction systems derived from a pretrained noise-robust automatic speech recognition (ASR) model for the second Clarity Prediction Challenge (CPC2). One system is intrusive and leverages the hidden representations of the ASR model. The other system is non-intrusive and makes predictions with derived ASR uncertainty. The ASR model is only pretrained with a simulated noisy speech corpus and does not take advantage of the CPC2 data. For that reason, the intelligibility prediction systems are robust to unseen scenarios given the accurate prediction performance on the CPC2 evaluation.

The Cadenza project aims to improve the audio quality of music for those who have a hearing loss. This is being done through a series of signal processing challenges, to foster better and more inclusive technologies. In the first round, two common listening scenarios are considered: listening to music over headphones, and with a hearing aid in a car. The first scenario is cast as a demixing-remixing problem, where the music is decomposed into vocals, bass, drums and other components. These can then be intelligently remixed in a personalized way, to increase the audio quality for a person who has a hearing loss. In the second scenario, music is coming from car loudspeakers, and the music has to be enhanced to overcome the masking effect of the car noise. This is done by taking into account the music, the hearing ability of the listener, the hearing aid and the speed of the car. The audio quality of the submissions will be evaluated using the Hearing Aid Audio Quality Index (HAAQI) for objective assessment and by a panel of people with hearing loss for subjective evaluation.

This paper reports on the design and results of the 2024 ICASSP SP Cadenza Challenge: Music Demixing/Remixing for Hearing Aids. The Cadenza project is working to enhance the audio quality of music for those with a hearing loss. The scenario for the challenge was listening to stereo reproduction over loudspeakers via hearing aids. The task was to: decompose pop/rock music into vocal, drums, bass and other (VDBO); rebalance the different tracks with specified gains and then remixing back to stereo. End-to-end approaches were also accepted. 17 systems were submitted by 11 teams. Causal systems performed poorer than non-causal approaches. 9 systems beat the baseline. A common approach was to fine-tuning pretrained demixing models. The best approach used an ensemble of models., Comment: 2-page paper for ICASSP 2024 SP Grand Challenge

Core-collapse supernovae (CCSNe) are potentially capable of producing large quantities of dust, with strong evidence that ejecta dust masses can grow significantly over extended periods of time. Red-blue asymmetries in the broad emission lines of CCSNe can be modelled using the Monte Carlo radiative transfer code DAMOCLES, to determine ejecta dust masses. To facilitate easier use of DAMOCLES, we present a Tkinter graphical user interface (GUI) running DAMOCLES. The GUI was tested by high school students as part of the Original Research By Young Twinkle Students (ORBYTS) programme, who used it to measure the dust masses formed at two epochs in two Type IIP CCSNe: SN 2012aw and iPTF14hls, demonstrating that a wide range of people can contribute significantly to scientific advancement. Bayesian methods were used to quantify uncertainties on our model parameters. From the presence of a red scattering wing in the day 1863 H$\alpha$ profile of SN 2012aw, we were able to constrain the dust composition to large (radius $>0.1 \mu$m) silicate grains, with a dust mass of $6.0^{+21.9}_{-3.6}\times10^{-4} M_\odot$. From the day 1158 H$\alpha$ profile of SN 2012aw, we found a dust mass of $3.0^{+14}_{-2.5}\times10^{-4}$ M$_\odot$. For iPTF14hls, we found a day 1170 dust mass of 8.1 $^{+81}_{-7.6}\times10^{-5}$ M$_{\odot}$ for a dust composition consisting of 50% amorphous carbon and 50% astronomical silicate. At 1000 days post explosion, SN 2012aw and iPTF14hls have formed less dust than SN 1987A, suggesting that SN 1987A could form larger dust masses than other Type IIP's., Comment: Accepted by MNRAS on 28/11/2022, 10 pages, 6 figures. Author accepted manuscript

In this paper, we explore an improved framework to train a monoaural neural enhancement model for robust speech recognition. The designed training framework extends the existing mixture invariant training criterion to exploit both unpaired clean speech and real noisy data. It is found that the unpaired clean speech is crucial to improve quality of separated speech from real noisy speech. The proposed method also performs remixing of processed and unprocessed signals to alleviate the processing artifacts. Experiments on the single-channel CHiME-3 real test sets show that the proposed method improves significantly in terms of speech recognition performance over the enhancement system trained either on the mismatched simulated data in a supervised fashion or on the matched real data in an unsupervised fashion. Between 16% and 39% relative WER reduction has been achieved by the proposed system compared to the unprocessed signal using end-to-end and hybrid acoustic models without retraining on distorted data., Comment: Accepted to INTERSPEECH 2022

Non-intrusive intelligibility prediction is important for its application in realistic scenarios, where a clean reference signal is difficult to access. The construction of many non-intrusive predictors require either ground truth intelligibility labels or clean reference signals for supervised learning. In this work, we leverage an unsupervised uncertainty estimation method for predicting speech intelligibility, which does not require intelligibility labels or reference signals to train the predictor. Our experiments demonstrate that the uncertainty from state-of-the-art end-to-end automatic speech recognition (ASR) models is highly correlated with speech intelligibility. The proposed method is evaluated on two databases and the results show that the unsupervised uncertainty measures of ASR models are more correlated with speech intelligibility from listening results than the predictions made by widely used intrusive methods., Comment: Accepted to INTERSPEECH2022

An accurate objective speech intelligibility prediction algorithms is of great interest for many applications such as speech enhancement for hearing aids. Most algorithms measures the signal-to-noise ratios or correlations between the acoustic features of clean reference signals and degraded signals. However, these hand-picked acoustic features are usually not explicitly correlated with recognition. Meanwhile, deep neural network (DNN) based automatic speech recogniser (ASR) is approaching human performance in some speech recognition tasks. This work leverages the hidden representations from DNN-based ASR as features for speech intelligibility prediction in hearing-impaired listeners. The experiments based on a hearing aid intelligibility database show that the proposed method could make better prediction than a widely used short-time objective intelligibility (STOI) based binaural measure., Comment: Accepted to INTERSPEECH2022

End-to-end models have achieved significant improvement on automatic speech recognition. One common method to improve performance of these models is expanding the data-space through data augmentation. Meanwhile, human auditory inspired front-ends have also demonstrated improvement for automatic speech recognisers. In this work, a well-verified auditory-based model, which can simulate various hearing abilities, is investigated for the purpose of data augmentation for end-to-end speech recognition. By introducing the auditory model into the data augmentation process, end-to-end systems are encouraged to ignore variation from the signal that cannot be heard and thereby focus on robust features for speech recognition. Two mechanisms in the auditory model, spectral smearing and loudness recruitment, are studied on the LibriSpeech dataset with a transformer-based end-to-end model. The results show that the proposed augmentation methods can bring statistically significant improvement on the performance of the state-of-the-art SpecAugment.

Video compression is a central feature of the modern internet powering technologies from social media to video conferencing. While video compression continues to mature, for many compression settings, quality loss is still noticeable. These settings nevertheless have important applications to the efficient transmission of videos over bandwidth constrained or otherwise unstable connections. In this work, we develop a deep learning architecture capable of restoring detail to compressed videos which leverages the underlying structure and motion information embedded in the video bitstream. We show that this improves restoration accuracy compared to prior compression correction methods and is competitive when compared with recent deep-learning-based video compression methods on rate-distortion while achieving higher throughput. Furthermore, we condition our model on quantization data which is readily available in the bitstream. This allows our single model to handle a variety of different compression quality settings which required an ensemble of models in prior work., Comment: WACV 2024

In this paper, we introduce a novel semi-supervised learning framework for end-to-end speech separation. The proposed method first uses mixtures of unseparated sources and the mixture invariant training (MixIT) criterion to train a teacher model. The teacher model then estimates separated sources that are used to train a student model with standard permutation invariant training (PIT). The student model can be fine-tuned with supervised data, i.e., paired artificial mixtures and clean speech sources, and further improved via model distillation. Experiments with single and multi channel mixtures show that the teacher-student training resolves the over-separation problem observed in the original MixIT method. Further, the semisupervised performance is comparable to a fully-supervised separation system trained using ten times the amount of supervised data., Comment: Accepted to Interspeech 2021

Current hearing aids normally provide amplification based on a general prescriptive fitting, and the benefits provided by the hearing aids vary among different listening environments despite the inclusion of noise suppression feature. Motivated by this fact, this paper proposes a data-driven machine learning technique to develop hearing aid fittings that are customised to speech in different noisy environments. A differentiable hearing loss model is proposed and used to optimise fittings with back-propagation. The customisation is reflected on the data of speech in different noise with also the consideration of noise suppression. The objective evaluation shows the advantages of optimised custom fittings over general prescriptive fittings., Comment: Accepted to Interspeech 2021

Hearing aids are expected to improve speech intelligibility for listeners with hearing impairment. An appropriate amplification fitting tuned for the listener's hearing disability is critical for good performance. The developments of most prescriptive fittings are based on data collected in subjective listening experiments, which are usually expensive and time-consuming. In this paper, we explore an alternative approach to finding the optimal fitting by introducing a hearing aid speech processing framework, in which the fitting is optimised in an automated way using an intelligibility objective function based on the HASPI physiological auditory model. The framework is fully differentiable, thus can employ the back-propagation algorithm for efficient, data-driven optimisation. Our initial objective experiments show promising results for noise-free speech amplification, where the automatically optimised processors outperform one of the well recognised hearing aid prescriptions., Comment: To appear at ICASSP 2021

In this paper, we ask whether vocal source features (pitch, shimmer, jitter, etc) can improve the performance of automatic sung speech recognition, arguing that conclusions previously drawn from spoken speech studies may not be valid in the sung speech domain. We first use a parallel singing/speaking corpus (NUS-48E) to illustrate differences in sung vs spoken voicing characteristics including pitch range, syllables duration, vibrato, jitter and shimmer. We then use this analysis to inform speech recognition experiments on the sung speech DSing corpus, using a state of the art acoustic model and augmenting conventional features with various voice source parameters. Experiments are run with three standard (increasingly large) training sets, DSing1 (15.1 hours), DSing3 (44.7 hours) and DSing30 (149.1 hours). Pitch combined with degree of voicing produces a significant decrease in WER from 38.1% to 36.7% when training with DSing1 however smaller decreases in WER observed when training with the larger more varied DSing3 and DSing30 sets were not seen to be statistically significant. Voicing quality characteristics did not improve recognition performance although analysis suggests that they do contribute to an improved discrimination between voiced/unvoiced phoneme pairs., Comment: Accepted to ICASSP 2021

In this paper, we present a novel multi-channel speech extraction system to simultaneously extract multiple clean individual sources from a mixture in noisy and reverberant environments. The proposed method is built on an improved multi-channel time-domain speech separation network which employs speaker embeddings to identify and extract multiple targets without label permutation ambiguity. To efficiently inform the speaker information to the extraction model, we propose a new speaker conditioning mechanism by designing an additional speaker branch for receiving external speaker embeddings. Experiments on 2-channel WHAMR! data show that the proposed system improves by 9% relative the source separation performance over a strong multi-channel baseline, and it increases the speech recognition accuracy by more than 16% relative over the same baseline., Comment: Accepted for ICASSP 2021

This paper introduces a new method for multi-channel time domain speech separation in reverberant environments. A fully-convolutional neural network structure has been used to directly separate speech from multiple microphone recordings, with no need of conventional spatial feature extraction. To reduce the influence of reverberation on spatial feature extraction, a dereverberation pre-processing method has been applied to further improve the separation performance. A spatialized version of wsj0-2mix dataset has been simulated to evaluate the proposed system. Both source separation and speech recognition performance of the separated signals have been evaluated objectively. Experiments show that the proposed fully-convolutional network improves the source separation metric and the word error rate (WER) by more than 13% and 50% relative, respectively, over a reference system with conventional features. Applying dereverberation as pre-processing to the proposed system can further reduce the WER by 29% relative using an acoustic model trained on clean and reverberated data., Comment: Presented at IEEE ICASSP 2020

In the Clarity project, we will run a series of machine learning challenges to revolutionise speech processing for hearing devices. Over five years, there will be three paired challenges. Each pair will consist of a competition focussed on hearing-device processing ("enhancement") and another focussed on speech perception modelling ("prediction"). The enhancement challenges will deliver new and improved approaches for hearing device signal processing for speech. The parallel prediction challenges will develop and improve methods for predicting speech intelligibility and quality for hearing impaired listeners. This Engineering and Physical Sciences Research Council (EPSRC) funded project involves researchers from the Universities of Sheffield, Salford, Nottingham and Cardiff in conjunction with the Hearing Industry Research Consortium, Action on Hearing Loss, Amazon, and Honda. To register interest in the challenges, go to www.claritychallenge.org., Comment: 3 pages, 2 figures

Following the success of the 1st, 2nd, 3rd, 4th and 5th CHiME challenges we organize the 6th CHiME Speech Separation and Recognition Challenge (CHiME-6). The new challenge revisits the previous CHiME-5 challenge and further considers the problem of distant multi-microphone conversational speech diarization and recognition in everyday home environments. Speech material is the same as the previous CHiME-5 recordings except for accurate array synchronization. The material was elicited using a dinner party scenario with efforts taken to capture data that is representative of natural conversational speech. This paper provides a baseline description of the CHiME-6 challenge for both segmented multispeaker speech recognition (Track 1) and unsegmented multispeaker speech recognition (Track 2). Of note, Track 2 is the first challenge activity in the community to tackle an unsegmented multispeaker speech recognition scenario with a complete set of reproducible open source baselines providing speech enhancement, speaker diarization, and speech recognition modules.

We present an approach for high-resolution video frame prediction by conditioning on both past frames and past optical flows. Previous approaches rely on resampling past frames, guided by a learned future optical flow, or on direct generation of pixels. Resampling based on flow is insufficient because it cannot deal with disocclusions. Generative models currently lead to blurry results. Recent approaches synthesis a pixel by convolving input patches with a predicted kernel. However, their memory requirement increases with kernel size. Here, we spatially-displaced convolution (SDC) module for video frame prediction. We learn a motion vector and a kernel for each pixel and synthesize a pixel by applying the kernel at a displaced location in the source image, defined by the predicted motion vector. Our approach inherits the merits of both vector-based and kernel-based approaches, while ameliorating their respective disadvantages. We train our model on 428K unlabelled 1080p video game frames. Our approach produces state-of-the-art results, achieving an SSIM score of 0.904 on high-definition YouTube-8M videos, 0.918 on Caltech Pedestrian videos. Our model handles large motion effectively and synthesizes crisp frames with consistent motion., Comment: Published in ECCV 2018. Codes available at https://github.com/NVIDIA/semantic-segmentation/tree/sdcnet/sdcnet. Project page available at https://nv-adlr.github.io/publication/2018-SDCNet

Human auditory cortex excels at selectively suppressing background noise to focus on a target speaker. The process of selective attention in the brain is known to contextually exploit the available audio and visual cues to better focus on target speaker while filtering out other noises. In this study, we propose a novel deep neural network (DNN) based audiovisual (AV) mask estimation model. The proposed AV mask estimation model contextually integrates the temporal dynamics of both audio and noise-immune visual features for improved mask estimation and speech separation. For optimal AV features extraction and ideal binary mask (IBM) estimation, a hybrid DNN architecture is exploited to leverages the complementary strengths of a stacked long short term memory (LSTM) and convolution LSTM network. The comparative simulation results in terms of speech quality and intelligibility demonstrate significant performance improvement of our proposed AV mask estimation model as compared to audio-only and visual-only mask estimation approaches for both speaker dependent and independent scenarios., Comment: Accepted for Interspeech 2018, 5 pages, 4 figures

The CHiME challenge series aims to advance robust automatic speech recognition (ASR) technology by promoting research at the interface of speech and language processing, signal processing , and machine learning. This paper introduces the 5th CHiME Challenge, which considers the task of distant multi-microphone conversational ASR in real home environments. Speech material was elicited using a dinner party scenario with efforts taken to capture data that is representative of natural conversational speech and recorded by 6 Kinect microphone arrays and 4 binaural microphone pairs. The challenge features a single-array track and a multiple-array track and, for each track, distinct rankings will be produced for systems focusing on robustness with respect to distant-microphone capture vs. systems attempting to address all aspects of the task including conversational language modeling. We discuss the rationale for the challenge and provide a detailed description of the data collection procedure, the task, and the baseline systems for array synchronization, speech enhancement, and conventional and end-to-end ASR.

In this work we introduce malware detection from raw byte sequences as a fruitful research area to the larger machine learning community. Building a neural network for such a problem presents a number of interesting challenges that have not occurred in tasks such as image processing or NLP. In particular, we note that detection from raw bytes presents a sequence problem with over two million time steps and a problem where batch normalization appear to hinder the learning process. We present our initial work in building a solution to tackle this problem, which has linear complexity dependence on the sequence length, and allows for interpretable sub-regions of the binary to be identified. In doing so we will discuss the many challenges in building a neural network to process data at this scale, and the methods we used to work around them.