Juan Pablo Bello

Chuyang Chen, Sivan Ding, Adrian S. Roman et al.

Pretrained spatial audio encoders are increasingly used as general-purpose representations for perceptual tasks, yet their spatial encoding capabilities remain poorly understood. We introduce the Spatial Audio Representation Learning (SARL) benchmark, a controlled framework for evaluating spatial information in pretrained audio models. SARL probes source-level factors (azimuth, elevation, distance, class) and room-level factors (RT60, volume, shape). Experiments across diverse encoders reveal three patterns: input configuration and training paradigm shape spatial encoding; source factors are consistently easier to decode than room factors; and sensitivity analysis under controlled perturbations shows heterogeneous responses to source and room variation. These results reveal systematic biases in current pretrained audio representations. SARL is released as an open-source benchmark for reproducible evaluation of spatial audio representations.

Sangeeta Srivastava, Ho-Hsiang Wu, Joao Rulff et al. · meta-ai

Audio applications involving environmental sound analysis increasingly use general-purpose audio representations, also known as embeddings, for transfer learning. Recently, Holistic Evaluation of Audio Representations (HEAR) evaluated twenty-nine embedding models on nineteen diverse tasks. However, the evaluation's effectiveness depends on the variation already captured within a given dataset. Therefore, for a given data domain, it is unclear how the representations would be affected by the variations caused by myriad microphones' range and acoustic conditions -- commonly known as channel effects. We aim to extend HEAR to evaluate invariance to channel effects in this work. To accomplish this, we imitate channel effects by injecting perturbations to the audio signal and measure the shift in the new (perturbed) embeddings with three distance measures, making the evaluation domain-dependent but not task-dependent. Combined with the downstream performance, it helps us make a more informed prediction of how robust the embeddings are to the channel effects. We evaluate two embeddings -- YAMNet, and OpenL3 on monophonic (UrbanSound8K) and polyphonic (SONYC-UST) urban datasets. We show that one distance measure does not suffice in such task-independent evaluation. Although Fréchet Audio Distance (FAD) correlates with the trend of the performance drop in the downstream task most accurately, we show that we need to study FAD in conjunction with the other distances to get a clear understanding of the overall effect of the perturbation. In terms of the embedding performance, we find OpenL3 to be more robust than YAMNet, which aligns with the HEAR evaluation.

Julia Wilkins, Justin Salamon, Magdalena Fuentes et al.

Finding the right sound effects (SFX) to match moments in a video is a difficult and time-consuming task, and relies heavily on the quality and completeness of text metadata. Retrieving high-quality (HQ) SFX using a video frame directly as the query is an attractive alternative, removing the reliance on text metadata and providing a low barrier to entry for non-experts. Due to the lack of HQ audio-visual training data, previous work on audio-visual retrieval relies on YouTube (in-the-wild) videos of varied quality for training, where the audio is often noisy and the video of amateur quality. As such it is unclear whether these systems would generalize to the task of matching HQ audio to production-quality video. To address this, we propose a multimodal framework for recommending HQ SFX given a video frame by (1) leveraging large language models and foundational vision-language models to bridge HQ audio and video to create audio-visual pairs, resulting in a highly scalable automatic audio-visual data curation pipeline; and (2) using pre-trained audio and visual encoders to train a contrastive learning-based retrieval system. We show that our system, trained using our automatic data curation pipeline, significantly outperforms baselines trained on in-the-wild data on the task of HQ SFX retrieval for video. Furthermore, while the baselines fail to generalize to this task, our system generalizes well from clean to in-the-wild data, outperforming the baselines on a dataset of YouTube videos despite only being trained on the HQ audio-visual pairs. A user study confirms that people prefer SFX retrieved by our system over the baseline 67% of the time both for HQ and in-the-wild data. Finally, we present ablations to determine the impact of model and data pipeline design choices on downstream retrieval performance. Please visit our project website to listen to and view our SFX retrieval results.

Rajsuryan Singh, Pablo Zinemanas, Xavier Serra et al.

Most recent work in visual sound source localization relies on semantic audio-visual representations learned in a self-supervised manner, and by design excludes temporal information present in videos. While it proves to be effective for widely used benchmark datasets, the method falls short for challenging scenarios like urban traffic. This work introduces temporal context into the state-of-the-art methods for sound source localization in urban scenes using optical flow as a means to encode motion information. An analysis of the strengths and weaknesses of our methods helps us better understand the problem of visual sound source localization and sheds light on open challenges for audio-visual scene understanding.

Julia Wilkins, Jaehun Kim, Matthew E. P. Davies et al.

Music representations are the backbone of modern recommendation systems, powering playlist generation, similarity search, and personalized discovery. Yet most embeddings offer little control for adjusting a single musical attribute, e.g., changing only the mood of a track while preserving its genre or instrumentation. In this work, we address the problem of controllable music retrieval through embedding-based transformation, where the objective is to retrieve songs that remain similar to a seed track but are modified along one chosen dimension. We propose a novel framework for mood-guided music embedding transformation, which learns a mapping from a seed audio embedding to a target embedding guided by mood labels, while preserving other musical attributes. Because mood cannot be directly altered in the seed audio, we introduce a sampling mechanism that retrieves proxy targets to balance diversity with similarity to the seed. We train a lightweight translation model using this sampling strategy and introduce a novel joint objective that encourages transformation and information preservation. Extensive experiments on two datasets show strong mood transformation performance while retaining genre and instrumentation far better than training-free baselines, establishing controllable embedding transformation as a promising paradigm for personalized music retrieval.

Ho-Hsiang Wu, Prem Seetharaman, Kundan Kumar et al.

We propose Wav2CLIP, a robust audio representation learning method by distilling from Contrastive Language-Image Pre-training (CLIP). We systematically evaluate Wav2CLIP on a variety of audio tasks including classification, retrieval, and generation, and show that Wav2CLIP can outperform several publicly available pre-trained audio representation algorithms. Wav2CLIP projects audio into a shared embedding space with images and text, which enables multimodal applications such as zero-shot classification, and cross-modal retrieval. Furthermore, Wav2CLIP needs just ~10% of the data to achieve competitive performance on downstream tasks compared with fully supervised models, and is more efficient to pre-train than competing methods as it does not require learning a visual model in concert with an auditory model. Finally, we demonstrate image generation from Wav2CLIP as qualitative assessment of the shared embedding space. Our code and model weights are open sourced and made available for further applications.

Yu Wang, Nicholas J. Bryan, Justin Salamon et al.

Few-shot learning aims to train models that can recognize novel classes given just a handful of labeled examples, known as the support set. While the field has seen notable advances in recent years, they have often focused on multi-class image classification. Audio, in contrast, is often multi-label due to overlapping sounds, resulting in unique properties such as polyphony and signal-to-noise ratios (SNR). This leads to unanswered questions concerning the impact such audio properties may have on few-shot learning system design, performance, and human-computer interaction, as it is typically up to the user to collect and provide inference-time support set examples. We address these questions through a series of experiments designed to elucidate the answers to these questions. We introduce two novel datasets, FSD-MIX-CLIPS and FSD-MIX-SED, whose programmatic generation allows us to explore these questions systematically. Our experiments lead to audio-specific insights on few-shot learning, some of which are at odds with recent findings in the image domain: there is no best one-size-fits-all model, method, and support set selection criterion. Rather, it depends on the expected application scenario. Our code and data are available at https://github.com/wangyu/rethink-audio-fsl.

Jong Wook Kim, Justin Salamon, Peter Li et al.

The task of estimating the fundamental frequency of a monophonic sound recording, also known as pitch tracking, is fundamental to audio processing with multiple applications in speech processing and music information retrieval. To date, the best performing techniques, such as the pYIN algorithm, are based on a combination of DSP pipelines and heuristics. While such techniques perform very well on average, there remain many cases in which they fail to correctly estimate the pitch. In this paper, we propose a data-driven pitch tracking algorithm, CREPE, which is based on a deep convolutional neural network that operates directly on the time-domain waveform. We show that the proposed model produces state-of-the-art results, performing equally or better than pYIN. Furthermore, we evaluate the model's generalizability in terms of noise robustness. A pre-trained version of CREPE is made freely available as an open-source Python module for easy application.

Magdalena Fuentes, Justin Salamon, Pablo Zinemanas et al.

Soundata is a Python library for loading and working with audio datasets in a standardized way, removing the need for writing custom loaders in every project, and improving reproducibility by providing tools to validate data against a canonical version. It speeds up research pipelines by allowing users to quickly download a dataset, load it into memory in a standardized and reproducible way, validate that the dataset is complete and correct, and more. Soundata is based and inspired on mirdata and design to complement mirdata by working with environmental sound, bioacoustic and speech datasets, among others. Soundata was created to be easy to use, easy to contribute to, and to increase reproducibility and standardize usage of sound datasets in a flexible way.

Aurora Cramer, Mark Cartwright, Fatemeh Pishdadian et al.

While the estimation of what sound sources are, when they occur, and from where they originate has been well-studied, the estimation of how loud these sound sources are has been often overlooked. Current solutions to this task, which we refer to as source-specific sound level estimation (SSSLE), suffer from challenges due to the impracticality of acquiring realistic data and a lack of robustness to realistic recording conditions. Recently proposed weakly supervised source separation offer a means of leveraging clip-level source annotations to train source separation models, which we augment with modified loss functions to bridge the gap between source separation and SSSLE and to address the presence of background. We show that our approach improves SSSLE performance compared to baseline source separation models and provide an ablation analysis to explore our method's design choices, showing that SSSLE in practical recording and annotation scenarios is possible.

Ho-Hsiang Wu, Chieh-Chi Kao, Qingming Tang et al.

Deep learning is very data hungry, and supervised learning especially requires massive labeled data to work well. Machine listening research often suffers from limited labeled data problem, as human annotations are costly to acquire, and annotations for audio are time consuming and less intuitive. Besides, models learned from labeled dataset often embed biases specific to that particular dataset. Therefore, unsupervised learning techniques become popular approaches in solving machine listening problems. Particularly, a self-supervised learning technique utilizing reconstructions of multiple hand-crafted audio features has shown promising results when it is applied to speech domain such as emotion recognition and automatic speech recognition (ASR). In this paper, we apply self-supervised and multi-task learning methods for pre-training music encoders, and explore various design choices including encoder architectures, weighting mechanisms to combine losses from multiple tasks, and worker selections of pretext tasks. We investigate how these design choices interact with various downstream music classification tasks. We find that using various music specific workers altogether with weighting mechanisms to balance the losses during pre-training helps improve and generalize to the downstream tasks.

Mark Cartwright, Jason Cramer, Ana Elisa Mendez Mendez et al.

We present SONYC-UST-V2, a dataset for urban sound tagging with spatiotemporal information. This dataset is aimed for the development and evaluation of machine listening systems for real-world urban noise monitoring. While datasets of urban recordings are available, this dataset provides the opportunity to investigate how spatiotemporal metadata can aid in the prediction of urban sound tags. SONYC-UST-V2 consists of 18510 audio recordings from the "Sounds of New York City" (SONYC) acoustic sensor network, including the timestamp of audio acquisition and location of the sensor. The dataset contains annotations by volunteers from the Zooniverse citizen science platform, as well as a two-stage verification with our team. In this article, we describe our data collection procedure and propose evaluation metrics for multilabel classification of urban sound tags. We report the results of a simple baseline model that exploits spatiotemporal information.

Yu Wang, Justin Salamon, Mark Cartwright et al.

Data-driven approaches to automatic drum transcription (ADT) are often limited to a predefined, small vocabulary of percussion instrument classes. Such models cannot recognize out-of-vocabulary classes nor are they able to adapt to finer-grained vocabularies. In this work, we address open vocabulary ADT by introducing few-shot learning to the task. We train a Prototypical Network on a synthetic dataset and evaluate the model on multiple real-world ADT datasets with polyphonic accompaniment. We show that, given just a handful of selected examples at inference time, we can match and in some cases outperform a state-of-the-art supervised ADT approach under a fixed vocabulary setting. At the same time, we show that our model can successfully generalize to finer-grained or extended vocabularies unseen during training, a scenario where supervised approaches cannot operate at all. We provide a detailed analysis of our experimental results, including a breakdown of performance by sound class and by polyphony.

Vincent Lostanlen, Alice Cohen-Hadria, Juan Pablo Bello

With the aim of constructing a biologically plausible model of machine listening, we study the representation of a multicomponent stationary signal by a wavelet scattering network. First, we show that renormalizing second-order nodes by their first-order parents gives a simple numerical criterion to assess whether two neighboring components will interfere psychoacoustically. Secondly, we run a manifold learning algorithm (Isomap) on scattering coefficients to visualize the similarity space underlying parametric additive synthesis. Thirdly, we generalize the "one or two components" framework to three sine waves or more, and prove that the effective scattering depth of a Fourier series grows in logarithmic proportion to its bandwidth.

Vincent Lostanlen, Kaitlin Palmer, Elly Knight et al.

This paper proposes to perform unsupervised detection of bioacoustic events by pooling the magnitudes of spectrogram frames after per-channel energy normalization (PCEN). Although PCEN was originally developed for speech recognition, it also has beneficial effects in enhancing animal vocalizations, despite the presence of atmospheric absorption and intermittent noise. We prove that PCEN generalizes logarithm-based spectral flux, yet with a tunable time scale for background noise estimation. In comparison with pointwise logarithm, PCEN reduces false alarm rate by 50x in the near field and 5x in the far field, both on avian and marine bioacoustic datasets. Such improvements come at moderate computational cost and require no human intervention, thus heralding a promising future for PCEN in bioacoustics.

Vincent Lostanlen, Sripathi Sridhar, Brian McFee et al.

To explain the consonance of octaves, music psychologists represent pitch as a helix where azimuth and axial coordinate correspond to pitch class and pitch height respectively. This article addresses the problem of discovering this helical structure from unlabeled audio data. We measure Pearson correlations in the constant-Q transform (CQT) domain to build a K-nearest neighbor graph between frequency subbands. Then, we run the Isomap manifold learning algorithm to represent this graph in a three-dimensional space in which straight lines approximate graph geodesics. Experiments on isolated musical notes demonstrate that the resulting manifold resembles a helix which makes a full turn at every octave. A circular shape is also found in English speech, but not in urban noise. We discuss the impact of various design choices on the visualization: instrumentarium, loudness mapping function, and number of neighbors K.

Jong Wook Kim, Juan Pablo Bello

Automatic music transcription is considered to be one of the hardest problems in music information retrieval, yet recent deep learning approaches have achieved substantial improvements on transcription performance. These approaches commonly employ supervised learning models that predict various time-frequency representations, by minimizing element-wise losses such as the cross entropy function. However, applying the loss in this manner assumes conditional independence of each label given the input, and thus cannot accurately express inter-label dependencies. To address this issue, we introduce an adversarial training scheme that operates directly on the time-frequency representations and makes the output distribution closer to the ground-truth. Through adversarial learning, we achieve a consistent improvement in both frame-level and note-level metrics over Onsets and Frames, a state-of-the-art music transcription model. Our results show that adversarial learning can significantly reduce the error rate while increasing the confidence of the model estimations. Our approach is generic and applicable to any transcription model based on multi-label predictions, which are very common in music signal analysis.

Vincent Lostanlen, Justin Salamon, Andrew Farnsworth et al.

Bioacoustic sensors, sometimes known as autonomous recording units (ARUs), can record sounds of wildlife over long periods of time in scalable and minimally invasive ways. Deriving per-species abundance estimates from these sensors requires detection, classification, and quantification of animal vocalizations as individual acoustic events. Yet, variability in ambient noise, both over time and across sensors, hinders the reliability of current automated systems for sound event detection (SED), such as convolutional neural networks (CNN) in the time-frequency domain. In this article, we develop, benchmark, and combine several machine listening techniques to improve the generalizability of SED models across heterogeneous acoustic environments. As a case study, we consider the problem of detecting avian flight calls from a ten-hour recording of nocturnal bird migration, recorded by a network of six ARUs in the presence of heterogeneous background noise. Starting from a CNN yielding state-of-the-art accuracy on this task, we introduce two noise adaptation techniques, respectively integrating short-term (60 milliseconds) and long-term (30 minutes) context. First, we apply per-channel energy normalization (PCEN) in the time-frequency domain, which applies short-term automatic gain control to every subband in the mel-frequency spectrogram. Secondly, we replace the last dense layer in the network by a context-adaptive neural network (CA-NN) layer. Combining them yields state-of-the-art results that are unmatched by artificial data augmentation alone. We release a pre-trained version of our best performing system under the name of BirdVoxDetect, a ready-to-use detector of avian flight calls in field recordings.

Charlie Mydlarz, Mohit Sharma, Yitzchak Lockerman et al.

Noise pollution is one of the topmost quality of life issues for urban residents in the United States. Continued exposure to high levels of noise has proven effects on health, including acute effects such as sleep disruption, and long-term effects such as hypertension, heart disease, and hearing loss. To investigate and ultimately aid in the mitigation of urban noise, a network of 55 sensor nodes has been deployed across New York City for over two years, collecting sound pressure level (SPL) and audio data. This network has cumulatively amassed over 75 years of calibrated, high-resolution SPL measurements and 35 years of audio data. In addition, high frequency telemetry data has been collected that provides an indication of a sensors' health. This telemetry data was analyzed over an 18 month period across 31 of the sensors. It has been used to develop a prototype model for pre-failure detection which has the ability to identify sensors in a prefail state 69.1% of the time. The entire network infrastructure is outlined, including the operation of the sensors, followed by an analysis of its data yield and the development of the fault detection approach and the future system integration plans for this.

Jong Wook Kim, Rachel Bittner, Aparna Kumar et al.

The recent success of raw audio waveform synthesis models like WaveNet motivates a new approach for music synthesis, in which the entire process --- creating audio samples from a score and instrument information --- is modeled using generative neural networks. This paper describes a neural music synthesis model with flexible timbre controls, which consists of a recurrent neural network conditioned on a learned instrument embedding followed by a WaveNet vocoder. The learned embedding space successfully captures the diverse variations in timbres within a large dataset and enables timbre control and morphing by interpolating between instruments in the embedding space. The synthesis quality is evaluated both numerically and perceptually, and an interactive web demo is presented.

Juan Pablo Bello, Claudio Silva, Oded Nov et al.

We present the Sounds of New York City (SONYC) project, a smart cities initiative focused on developing a cyber-physical system for the monitoring, analysis and mitigation of urban noise pollution. Noise pollution is one of the topmost quality of life issues for urban residents in the U.S. with proven effects on health, education, the economy, and the environment. Yet, most cities lack the resources to continuously monitor noise and understand the contribution of individual sources, the tools to analyze patterns of noise pollution at city-scale, and the means to empower city agencies to take effective, data-driven action for noise mitigation. The SONYC project advances novel technological and socio-technical solutions that help address these needs. SONYC includes a distributed network of both sensors and people for large-scale noise monitoring. The sensors use low-cost, low-power technology, and cutting-edge machine listening techniques, to produce calibrated acoustic measurements and recognize individual sound sources in real time. Citizen science methods are used to help urban residents connect to city agencies and each other, understand their noise footprint, and facilitate reporting and self-regulation. Crucially, SONYC utilizes big data solutions to analyze, retrieve and visualize information from sensors and citizens, creating a comprehensive acoustic model of the city that can be used to identify significant patterns of noise pollution. These data can be used to drive the strategic application of noise code enforcement by city agencies to optimize the reduction of noise pollution. The entire system, integrating cyber, physical and social infrastructure, forms a closed loop of continuous sensing, analysis and actuation on the environment. SONYC provides a blueprint for the mitigation of noise pollution that can potentially be applied to other cities in the US and abroad.

Brian McFee, Justin Salamon, Juan Pablo Bello

Sound event detection (SED) methods are tasked with labeling segments of audio recordings by the presence of active sound sources. SED is typically posed as a supervised machine learning problem, requiring strong annotations for the presence or absence of each sound source at every time instant within the recording. However, strong annotations of this type are both labor- and cost-intensive for human annotators to produce, which limits the practical scalability of SED methods. In this work, we treat SED as a multiple instance learning (MIL) problem, where training labels are static over a short excerpt, indicating the presence or absence of sound sources but not their temporal locality. The models, however, must still produce temporally dynamic predictions, which must be aggregated (pooled) when comparing against static labels during training. To facilitate this aggregation, we develop a family of adaptive pooling operators---referred to as auto-pool---which smoothly interpolate between common pooling operators, such as min-, max-, or average-pooling, and automatically adapt to the characteristics of the sound sources in question. We evaluate the proposed pooling operators on three datasets, and demonstrate that in each case, the proposed methods outperform non-adaptive pooling operators for static prediction, and nearly match the performance of models trained with strong, dynamic annotations. The proposed method is evaluated in conjunction with convolutional neural networks, but can be readily applied to any differentiable model for time-series label prediction.

Justin Salamon, Juan Pablo Bello

The ability of deep convolutional neural networks (CNN) to learn discriminative spectro-temporal patterns makes them well suited to environmental sound classification. However, the relative scarcity of labeled data has impeded the exploitation of this family of high-capacity models. This study has two primary contributions: first, we propose a deep convolutional neural network architecture for environmental sound classification. Second, we propose the use of audio data augmentation for overcoming the problem of data scarcity and explore the influence of different augmentations on the performance of the proposed CNN architecture. Combined with data augmentation, the proposed model produces state-of-the-art results for environmental sound classification. We show that the improved performance stems from the combination of a deep, high-capacity model and an augmented training set: this combination outperforms both the proposed CNN without augmentation and a "shallow" dictionary learning model with augmentation. Finally, we examine the influence of each augmentation on the model's classification accuracy for each class, and observe that the accuracy for each class is influenced differently by each augmentation, suggesting that the performance of the model could be improved further by applying class-conditional data augmentation.

Charlie Mydlarz, Justin Salamon, Juan Pablo Bello

The urban sound environment of New York City (NYC) can be, amongst other things: loud, intrusive, exciting and dynamic. As indicated by the large majority of noise complaints registered with the NYC 311 information/complaints line, the urban sound environment has a profound effect on the quality of life of the city's inhabitants. To monitor and ultimately understand these sonic environments, a process of long-term acoustic measurement and analysis is required. The traditional method of environmental acoustic monitoring utilizes short term measurement periods using expensive equipment, setup and operated by experienced and costly personnel. In this paper a different approach is proposed to this application which implements a smart, low-cost, static, acoustic sensing device based around consumer hardware. These devices can be deployed in numerous and varied urban locations for long periods of time, allowing for the collection of longitudinal urban acoustic data. The varied environmental conditions of urban settings make for a challenge in gathering calibrated sound pressure level data for prospective stakeholders. This paper details the sensors' design, development and potential future applications, with a focus on the calibration of the devices' Microelectromechanical systems (MEMS) microphone in order to generate reliable decibel levels at the type/class 2 level.