Spatial sound reasoning is a fundamental human skill, enabling us to navigate and interpret our surroundings based on sound. In this paper we present BAT, which combines the spatial sound perception ability of a binaural acoustic scene analysis model with the natural language reasoning capabilities of a large language model (LLM) to replicate this innate ability. To address the lack of existing datasets of in-the-wild spatial sounds, we synthesized a binaural audio dataset using AudioSet and SoundSpaces 2.0. Next, we developed SpatialSoundQA, a spatial sound-based question-answering dataset, offering a range of QA tasks that train BAT in various aspects of spatial sound perception and reasoning. The acoustic front end encoder of BAT is a novel spatial audio encoder named Spatial Audio Spectrogram Transformer, or Spatial-AST, which by itself achieves strong performance across sound event detection, spatial localization, and distance estimation. By integrating Spatial-AST with LLaMA-2 7B model, BAT transcends standard Sound Event Localization and Detection (SELD) tasks, enabling the model to reason about the relationships between the sounds in its environment. Our experiments demonstrate BAT's superior performance on both spatial sound perception and reasoning, showcasing the immense potential of LLMs in navigating and interpreting complex spatial audio environments.
This technical report details our work towards building an enhanced audio-visual sound event localization and detection (SELD) network. We build on top of the audio-only SELDnet23 model and adapt it to be audio-visual by merging both audio and video information prior to the gated recurrent unit (GRU) of the audio-only network. Our model leverages YOLO and DETIC object detectors. We also build a framework that implements audio-visual data augmentation and audio-visual synthetic data generation. We deliver an audio-visual SELDnet system that outperforms the existing audio-visual SELD baseline.
Sound event localization and detection (SELD) is an important task in machine listening. Major advancements rely on simulated data with sound events in specific rooms and strong spatio-temporal labels. SELD data is simulated by convolving spatialy-localized room impulse responses (RIRs) with sound waveforms to place sound events in a soundscape. However, RIRs require manual collection in specific rooms. We present SpatialScaper, a library for SELD data simulation and augmentation. Compared to existing tools, SpatialScaper emulates virtual rooms via parameters such as size and wall absorption. This allows for parameterized placement (including movement) of foreground and background sound sources. SpatialScaper also includes data augmentation pipelines that can be applied to existing SELD data. As a case study, we use SpatialScaper to add rooms to the DCASE SELD data. Training a model with our data led to progressive performance improves as a direct function of acoustic diversity. These results show that SpatialScaper is valuable to train robust SELD models.
Environment shifts and conflicts present significant challenges for learning-based sound event localization and detection (SELD) methods. SELD systems, when trained in particular acoustic settings, often show restricted generalization capabilities for diverse acoustic environments. Furthermore, obtaining annotated samples for spatial sound events is notably costly. Deploying a SELD system in a new environment requires extensive time for re-training and fine-tuning. To overcome these challenges, we propose environment-adaptive Meta-SELD, designed for efficient adaptation to new environments using minimal data. Our method specifically utilizes computationally synthesized spatial data and employs Model-Agnostic Meta-Learning (MAML) on a pre-trained, environment-independent model. The method then utilizes fast adaptation to unseen real-world environments using limited samples from the respective environments. Inspired by the Learning-to-Forget approach, we introduce the concept of selective memory as a strategy for resolving conflicts across environments. This approach involves selectively memorizing target-environment-relevant information and adapting to the new environments through the selective attenuation of model parameters. In addition, we introduce environment representations to characterize different acoustic settings, enhancing the adaptability of our attenuation approach to various environments. We evaluate our proposed method on the development set of the Sony-TAu Realistic Spatial Soundscapes 2023 (STARSS23) dataset and computationally synthesized scenes. Experimental results demonstrate the superior performance of the proposed method compared to conventional supervised learning methods, particularly in localization.
Sound event localization and detection (SELD) is a task for the classification of sound events and the localization of direction of arrival (DoA) utilizing multichannel acoustic signals. Prior studies employ spectral and channel information as the embedding for temporal attention. However, this usage limits the deep neural network from extracting meaningful features from the spectral or spatial domains. Therefore, our investigation in this paper presents a novel framework termed the Channel-Spectro-Temporal Transformer (CST-former) that bolsters SELD performance through the independent application of attention mechanisms to distinct domains. The CST-former architecture employs distinct attention mechanisms to independently process channel, spectral, and temporal information. In addition, we propose an unfolded local embedding (ULE) technique for channel attention (CA) to generate informative embedding vectors including local spectral and temporal information. Empirical validation through experimentation on the 2022 and 2023 DCASE Challenge task3 datasets affirms the efficacy of employing attention mechanisms separated across each domain and the benefit of ULE, in enhancing SELD performance.
We propose a novel multimodal video benchmark - the Perception Test - to evaluate the perception and reasoning skills of pre-trained multimodal models (e.g. Flamingo, SeViLA, or GPT-4). Compared to existing benchmarks that focus on computational tasks (e.g. classification, detection or tracking), the Perception Test focuses on skills (Memory, Abstraction, Physics, Semantics) and types of reasoning (descriptive, explanatory, predictive, counterfactual) across video, audio, and text modalities, to provide a comprehensive and efficient evaluation tool. The benchmark probes pre-trained models for their transfer capabilities, in a zero-shot / few-shot or limited finetuning regime. For these purposes, the Perception Test introduces 11.6k real-world videos, 23s average length, designed to show perceptually interesting situations, filmed by around 100 participants worldwide. The videos are densely annotated with six types of labels (multiple-choice and grounded video question-answers, object and point tracks, temporal action and sound segments), enabling both language and non-language evaluations. The fine-tuning and validation splits of the benchmark are publicly available (CC-BY license), in addition to a challenge server with a held-out test split. Human baseline results compared to state-of-the-art video QA models show a substantial gap in performance (91.4% vs 46.2%), suggesting that there is significant room for improvement in multimodal video understanding. Dataset, baseline code, and challenge server are available at https://github.com/deepmind/perception_test
Sound event localization and detection (SELD) combines two subtasks: sound event detection (SED) and direction of arrival (DOA) estimation. SELD is usually tackled as an audio-only problem, but visual information has been recently included. Few audio-visual (AV)-SELD works have been published and most employ vision via face/object bounding boxes, or human pose keypoints. In contrast, we explore the integration of audio and visual feature embeddings extracted with pre-trained deep networks. For the visual modality, we tested ResNet50 and Inflated 3D ConvNet (I3D). Our comparison of AV fusion methods includes the AV-Conformer and Cross-Modal Attentive Fusion (CMAF) model. Our best models outperform the DCASE 2023 Task3 audio-only and AV baselines by a wide margin on the development set of the STARSS23 dataset, making them competitive amongst state-of-the-art results of the AV challenge, without model ensembling, heavy data augmentation, or prediction post-processing. Such techniques and further pre-training could be applied as next steps to improve performance.
Sound Event Detection and Localization (SELD) constitutes a complex task that depends on extensive multichannel audio recordings with annotated sound events and their respective locations. In this paper, we introduce a self-supervised approach for SELD adapted from the pre-training methodology of wav2vec 2.0, which learns representations directly from raw audio data, eliminating the need for supervision. By applying this approach to SELD, we can leverage a substantial amount of unlabeled 3D audio data to learn robust representations of sound events and their locations. Our method comprises two primary stages: pre-training and fine-tuning. In the pre-training phase, unlabeled 3D audio datasets are utilized to train our w2v-SELD model, capturing intricate high-level features and contextual information inherent in audio signals. Subsequently, in the fine-tuning stage, a smaller dataset with labeled SELD data fine-tunes the pre-trained model. Experimental results on benchmark datasets demonstrate the effectiveness of the proposed self-supervised approach for SELD. The model surpasses baseline systems provided with the datasets and achieves competitive performance comparable to state-of-the-art supervised methods. The code and pre-trained parameters of our w2v-SELD model are available in this repository.
While direction of arrival (DOA) of sound events is generally estimated from multichannel audio data recorded in a microphone array, sound events usually derive from visually perceptible source objects, e.g., sounds of footsteps come from the feet of a walker. This paper proposes an audio-visual sound event localization and detection (SELD) task, which uses multichannel audio and video information to estimate the temporal activation and DOA of target sound events. Audio-visual SELD systems can detect and localize sound events using signals from a microphone array and audio-visual correspondence. We also introduce an audio-visual dataset, Sony-TAu Realistic Spatial Soundscapes 2023 (STARSS23), which consists of multichannel audio data recorded with a microphone array, video data, and spatiotemporal annotation of sound events. Sound scenes in STARSS23 are recorded with instructions, which guide recording participants to ensure adequate activity and occurrences of sound events. STARSS23 also serves human-annotated temporal activation labels and human-confirmed DOA labels, which are based on tracking results of a motion capture system. Our benchmark results demonstrate the benefits of using visual object positions in audio-visual SELD tasks. The data is available at https://zenodo.org/record/7880637.
This study addresses the application of deep learning techniques in joint sound signal classification and localization networks. Current state-of-the-art sound source localization deep learning networks lack feature aggregation within their architecture. Feature aggregation enhances model performance by enabling the consolidation of information from different feature scales, thereby improving feature robustness and invariance. This is particularly important in SSL networks, which must differentiate direct and indirect acoustic signals. To address this gap, we adapt feature aggregation techniques from computer vision neural networks to signal detection neural networks. Additionally, we propose the Scale Encoding Network (SEN) for feature aggregation to encode features from various scales, compressing the network for more computationally efficient aggregation. To evaluate the efficacy of feature aggregation in SSL networks, we integrated the following computer vision feature aggregation sub-architectures into a SSL control architecture: Path Aggregation Network (PANet), Weighted Bi-directional Feature Pyramid Network (BiFPN), and SEN. These sub-architectures were evaluated using two metrics for signal classification and two metrics for direction-of-arrival regression. PANet and BiFPN are established aggregators in computer vision models, while the proposed SEN is a more compact aggregator. The results suggest that models incorporating feature aggregations outperformed the control model, the Sound Event Localization and Detection network (SELDnet), in both sound signal classification and localization. The feature aggregation techniques enhance the performance of sound detection neural networks, particularly in direction-of-arrival regression.
Sound event localization and detection (SELD) involves sound event detection (SED) and direction of arrival (DoA) estimation tasks. SED mainly relies on temporal dependencies to distinguish different sound classes, while DoA estimation depends on spatial correlations to estimate source directions. This paper addresses the need to simultaneously extract spatial-temporal information in audio signals to improve SELD performance. A novel block, the sliding-window graph-former (SwG-former), is designed to learn temporal context information of sound events based on their spatial correlations. The SwG-former block transforms audio signals into a graph representation and constructs graph vertices to capture higher abstraction levels for spatial correlations. It uses different-sized sliding windows to adapt various sound event durations and aggregates temporal features with similar spatial information while incorporating multi-head self-attention (MHSA) to model global information. Furthermore, as the cornerstone of message passing, a robust Conv2dAgg function is proposed and embedded into the block to aggregate the features of neighbor vertices. As a result, a SwG-former model, which stacks the SwG-former blocks, demonstrates superior performance compared to recent advanced SELD models. The SwG-former block is also integrated into the event-independent network version 2 (EINV2), called SwG-EINV2, which surpasses the state-of-the-art (SOTA) methods under the same acoustic environment.
As deeper and more complex models are developed for the task of sound event localization and detection (SELD), the demand for annotated spatial audio data continues to increase. Annotating field recordings with 360$^{\circ}$ video takes many hours from trained annotators, while recording events within motion-tracked laboratories are bounded by cost and expertise. Because of this, localization models rely on a relatively limited amount of spatial audio data in the form of spatial room impulse response (SRIR) datasets, which limits the progress of increasingly deep neural network based approaches. In this work, we demonstrate that simulated geometrical acoustics can provide an appealing solution to this problem. We use simulated geometrical acoustics to generate a novel SRIR dataset that can train a SELD model to provide similar performance to that of a real SRIR dataset. Furthermore, we demonstrate using simulated data to augment existing datasets, improving on benchmarks set by state of the art SELD models. We explore the potential and limitations of geometric acoustic simulation for localization and event detection. We also propose further studies to verify the limitations of this method, as well as further methods to generate synthetic data for SELD tasks without the need to record more data.
For learning-based sound event localization and detection (SELD) methods, different acoustic environments in the training and test sets may result in large performance differences in the validation and evaluation stages. Different environments, such as different sizes of rooms, different reverberation times, and different background noise, may be reasons for a learning-based system to fail. On the other hand, acquiring annotated spatial sound event samples, which include onset and offset time stamps, class types of sound events, and direction-of-arrival (DOA) of sound sources is very expensive. In addition, deploying a SELD system in a new environment often poses challenges due to time-consuming training and fine-tuning processes. To address these issues, we propose Meta-SELD, which applies meta-learning methods to achieve fast adaptation to new environments. More specifically, based on Model Agnostic Meta-Learning (MAML), the proposed Meta-SELD aims to find good meta-initialized parameters to adapt to new environments with only a small number of samples and parameter updating iterations. We can then quickly adapt the meta-trained SELD model to unseen environments. Our experiments compare fine-tuning methods from pre-trained SELD models with our Meta-SELD on the Sony-TAU Realistic Spatial Soundscapes 2023 (STARSSS23) dataset. The evaluation results demonstrate the effectiveness of Meta-SELD when adapting to new environments.
DNN-based methods have shown high performance in sound event localization and detection(SELD). While in real spatial sound scenes, reverberation and the imbalanced presence of various sound events increase the complexity of the SELD task. In this paper, we propose an effective SELD system in real spatial scenes.In our approach, a dynamic kernel convolution module is introduced after the convolution blocks to adaptively model the channel-wise features with different receptive fields. Secondly, we incorporate the SELDnet and EINv2 framework into the proposed SELD system with multi-track ACCDOA. Moreover, two scene-dedicated strategies are introduced into the training stage to improve the generalization of the system in realistic spatial sound scenes. Finally, we apply data augmentation methods to extend the dataset using channel rotation, spatial data synthesis. Four joint metrics are used to evaluate the performance of the SELD system on the Sony-TAu Realistic Spatial Soundscapes 2022 dataset.Experimental results show that the proposed systems outperform the fixed-kernel convolution SELD systems. In addition, the proposed system achieved an SELD score of 0.348 in the DCASE SELD task and surpassed the SOTA methods.
Localizing sounds and detecting events in different room environments is a difficult task, mainly due to the wide range of reflections and reverberations. When training neural network models with sounds recorded in only a few room environments, there is a tendency for the models to become overly specialized to those specific environments, resulting in overfitting. To address this overfitting issue, we propose divided spectro-temporal attention. In comparison to the baseline method, which utilizes a convolutional recurrent neural network (CRNN) followed by a temporal multi-head self-attention layer (MHSA), we introduce a separate spectral attention layer that aggregates spectral features prior to the temporal MHSA. To achieve efficient spectral attention, we reduce the frequency pooling size in the convolutional encoder of the baseline to obtain a 3D tensor that incorporates information about frequency, time, and channel. As a result, we can implement spectral attention with channel embeddings, which is not possible in the baseline method dealing with only temporal context in the RNN and MHSA layers. We demonstrate that the proposed divided spectro-temporal attention significantly improves the performance of sound event detection and localization scores for real test data from the STARSS23 development dataset. Additionally, we show that various data augmentations, such as frameshift, time masking, channel swapping, and moderate mix-up, along with the use of external data, contribute to the overall improvement in SELD performance.
Sound event localization and detection (SELD) combines the identification of sound events with the corresponding directions of arrival (DOA). Recently, event-oriented track output formats have been adopted to solve this problem; however, they still have limited generalization toward real-world problems in an unknown polyphony environment. To address the issue, we proposed an angular-distance-based multiple SELD (AD-YOLO), which is an adaptation of the "You Only Look Once" algorithm for SELD. The AD-YOLO format allows the model to learn sound occurrences location-sensitively by assigning class responsibility to DOA predictions. Hence, the format enables the model to handle the polyphony problem, regardless of the number of sound overlaps. We evaluated AD-YOLO on DCASE 2020-2022 challenge Task 3 datasets using four SELD objective metrics. The experimental results show that AD-YOLO achieved outstanding performance overall and also accomplished robustness in class-homogeneous polyphony environments.
In this article, we describe Conditioned Localizer and Classifier (CoLoC) which is a novel solution for Sound Event Localization and Detection (SELD). The solution constitutes of two stages: the localization is done first and is followed by classification conditioned by the output of the localizer. In order to resolve the problem of the unknown number of sources we incorporate the idea borrowed from Sequential Set Generation (SSG). Models from both stages are SELDnet-like CRNNs, but with single outputs. Conducted reasoning shows that such two single-output models are fit for SELD task. We show that our solution improves on the baseline system in most metrics on the STARSS22 Dataset.
Sound event localization and detection (SELD) is a joint task of sound event detection and direction-of-arrival estimation. In DCASE 2022 Task 3, types of data transform from computationally generated spatial recordings to recordings of real-sound scenes. Our system submitted to the DCASE 2022 Task 3 is based on our previous proposed Event-Independent Network V2 (EINV2) with a novel data augmentation method. Our method employs EINV2 with a track-wise output format, permutation-invariant training, and a soft parameter-sharing strategy, to detect different sound events of the same class but in different locations. The Conformer structure is used for extending EINV2 to learn local and global features. A data augmentation method, which contains several data augmentation chains composed of stochastic combinations of several different data augmentation operations, is utilized to generalize the model. To mitigate the lack of real-scene recordings in the development dataset and the presence of sound events being unbalanced, we exploit FSD50K, AudioSet, and TAU Spatial Room Impulse Response Database (TAU-SRIR DB) to generate simulated datasets for training. We present results on the validation set of Sony-TAu Realistic Spatial Soundscapes 2022 (STARSS22) in detail. Experimental results indicate that the ability to generalize to different environments and unbalanced performance among different classes are two main challenges. We evaluate our proposed method in Task 3 of the DCASE 2022 challenge and obtain the second rank in the teams ranking. Source code is released.
Performance of sound event localization and detection (SELD) in real scenes is limited by small size of SELD dataset, due to difficulty in obtaining sufficient amount of realistic multi-channel audio data recordings with accurate label. We used two main strategies to solve problems arising from the small real SELD dataset. First, we applied various data augmentation methods on all data dimensions: channel, frequency and time. We also propose original data augmentation method named Moderate Mixup in order to simulate situations where noise floor or interfering events exist. Second, we applied Squeeze-and-Excitation block on channel and frequency dimensions to efficiently extract feature characteristics. Result of our trained models on the STARSS22 test dataset achieved the best ER, F1, LE, and LR of 0.53, 49.8%, 16.0deg., and 56.2% respectively.
Spiking neural networks (SNNs) have demonstrated excellent capabilities in various intelligent scenarios. Most existing methods for training SNNs are based on the concept of synaptic plasticity; however, learning in the realistic brain also utilizes intrinsic non-synaptic mechanisms of neurons. The spike threshold of biological neurons is a critical intrinsic neuronal feature that exhibits rich dynamics on a millisecond timescale and has been proposed as an underlying mechanism that facilitates neural information processing. In this study, we develop a novel synergistic learning approach that involves simultaneously training synaptic weights and spike thresholds in SNNs. SNNs trained with synapse-threshold synergistic learning~(STL-SNNs) achieve significantly superior performance on various static and neuromorphic datasets than SNNs trained with two degenerated single-learning models. During training, the synergistic learning approach optimizes neural thresholds, providing the network with stable signal transmission via appropriate firing rates. Further analysis indicates that STL-SNNs are robust to noisy data and exhibit low energy consumption for deep network structures. Additionally, the performance of STL-SNN can be further improved by introducing a generalized joint decision framework. Overall, our findings indicate that biologically plausible synergies between synaptic and intrinsic non-synaptic mechanisms may provide a promising approach for developing highly efficient SNN learning methods.