Shengchen Li

Xubo Liu, Qiushi Huang, Xinhao Mei et al.

Audio captioning aims to generate text descriptions of audio clips. In the real world, many objects produce similar sounds. How to accurately recognize ambiguous sounds is a major challenge for audio captioning. In this work, inspired by inherent human multimodal perception, we propose visually-aware audio captioning, which makes use of visual information to help the description of ambiguous sounding objects. Specifically, we introduce an off-the-shelf visual encoder to extract video features and incorporate the visual features into an audio captioning system. Furthermore, to better exploit complementary audio-visual contexts, we propose an audio-visual attention mechanism that adaptively integrates audio and visual context and removes the redundant information in the latent space. Experimental results on AudioCaps, the largest audio captioning dataset, show that our proposed method achieves state-of-the-art results on machine translation metrics.

Yuanbo Hou, Yanru Wu, Qiaoqiao Ren et al.

Environmental sound understanding in computational auditory scene analysis (CASA) is often formulated as an audio-only recognition problem. This formulation leaves a persistent drawback in multi-label audio tagging (AT): acoustic similarity can make certain events difficult to separate from waveforms alone. In such cases, disambiguating cues often lie outside the waveform. Geospatial semantic context (GSC), derived from geographic information system data, e.g., points of interest (POI), provides location-tied environmental priors that can help reduce this ambiguity. A systematic study of this direction is enabled through the proposed geospatial audio tagging (Geo-AT) task, which conditions multi-label sound event tagging on GSC alongside audio. To benchmark Geo-AT, Geo-ATBench is introduced as a polyphonic audio benchmark with geographical annotations, containing 10.71 hours of audio across 28 event categories; each clip is paired with a GSC representation from 11 semantic context categories. GeoFusion-AT is proposed as a unified geo-audio fusion framework that evaluates feature-, representation-, and decision-level fusion on representative audio backbones, with audio- and GSC-only baselines. Results show that incorporating GSC improves AT performance, especially on acoustically confounded labels, indicating geospatial semantics provide effective priors beyond audio alone. A crowdsourced listening study with 10 participants on 579 samples shows that there is no significant difference in performance between models on Geo-ATBench labels and aggregated human labels, supporting Geo-ATBench as a human-aligned benchmark. The Geo-AT task, benchmark Geo-ATBench, and reproducible geo-audio fusion framework GeoFusion-AT provide a foundation for studying AT with geospatial semantic context within the CASA community. Dataset, code, models are on homepage (https://github.com/WuYanru2002/Geo-ATBench).

Jingyi Wang, Shengchen Li

Keyword spotting (KWS) is beneficial for voice-based user interactions with low-power devices at the edge. The edge devices are usually always-on, so edge computing brings bandwidth savings and privacy protection. The devices typically have limited memory spaces, computational performances, power and costs, for example, Cortex-M based microcontrollers. The challenge is to meet the high computation and low-latency requirements of deep learning on these devices. This paper firstly shows our small-footprint KWS system running on STM32F7 microcontroller with Cortex-M7 core @216MHz and 512KB static RAM. Our selected convolutional neural network (CNN) architecture has simplified number of operations for KWS to meet the constraint of edge devices. Our baseline system generates classification results for each 37ms including real-time audio feature extraction part. This paper further evaluates the actual performance for different pruning and quantization methods on microcontroller, including different granularity of sparsity, skipping zero weights, weight-prioritized loop order, and SIMD instruction. The result shows that for microcontrollers, there are considerable challenges for accelerate unstructured pruned models, and the structured pruning is more friendly than unstructured pruning. The result also verified that the performance improvement for quantization and SIMD instruction.

Yuqiang Li, Shengchen Li, George Fazekas

Pitch and meter are two fundamental music features for symbolic music generation tasks, where researchers usually choose different encoding methods depending on specific goals. However, the advantages and drawbacks of different encoding methods have not been frequently discussed. This paper presents a integrated analysis of the influence of two low-level feature, pitch and meter, on the performance of a token-based sequential music generation model. First, the commonly used MIDI number encoding and a less used class-octave encoding are compared. Second, an dense intra-bar metric grid is imposed to the encoded sequence as auxiliary features. Different complexity and resolutions of the metric grid are compared. For complexity, the single token approach and the multiple token approach are compared; for grid resolution, 0 (ablation), 1 (bar-level), 4 (downbeat-level) 12, (8th-triplet-level) up to 64 (64th-note-grid-level) are compared; for duration resolution, 4, 8, 12 and 16 subdivisions per beat are compared. All different encodings are tested on separately trained Transformer-XL models for a melody generation task. Regarding distribution similarity of several objective evaluation metrics to the test dataset, results suggest that the class-octave encoding significantly outperforms the taken-for-granted MIDI encoding on pitch-related metrics; finer grids and multiple-token grids improve the rhythmic quality, but also suffer from over-fitting at early training stage. Results display a general phenomenon of over-fitting from two aspects, the pitch embedding space and the test loss of the single-token grid encoding. From a practical perspective, we both demonstrate the feasibility and raise the concern of easy over-fitting problem of using smaller networks and lower embedding dimensions on the generation task. The findings can also contribute to futural models in terms of feature engineering.

Ruowei Xing, Shengchen Li

This paper refers to the extraction of multiple fundamental frequencies (multiple F0) based on PYIN, an algorithm for extracting the fundamental frequency (F0) of monophonic music, and a trained convolutional neural networks (CNN) model, where a pitch salience function of the input signal is produced to estimate the multiple F0. The implementation of these two algorithms and their corresponding advantages and disadvantages are discussed in this article. Analysing the different performance of these two methods, PYIN is applied to supplement the F0 extracted from the trained CNN model to combine the advantages of these two algorithms. For evaluation, four pieces played by two violins are used, and the performance of the models are evaluated accoring to the flatness of the F0 curve extracted. The result shows the combined model outperforms the original algorithms when extracting F0 from monophonic music and polyphonic music.

Leduo Chen, Junchuan Zhao, Shengchen Li

Timbre transfer aims to modify the timbral identity of a musical recording while preserving the original melody and rhythm. While single-instrument timbre transfer has made substantial progress, existing approaches to multi-instrument settings rely on separate-then-transfer pipelines that propagate source separation artifacts and produce incoherent synthesized timbres across stems. This paper proposes MixtureTT, to the best of our knowledge the first system for flexible per-stem timbre transfer directly from a polyphonic mixture. Given a mixture and a separate timbre reference for each target voice, MixtureTT jointly transfers all stems to the specified instruments through a shared diffusion process. Modeling the dependencies across the per-stem content and cross-stem harmonic, the proposed joint stem diffusion transformer eliminates cascaded separation error, reduces inference cost by a factor equal to the number of stems, and yields more coherent multi-stem outputs. Despite operating under a strictly harder input condition, evaluations on the SATB choral dataset show that MixtureTT outperforms single-instrument baselines on both objective and subjective metrics demonstrating the necessity of dedicated multi-instrument timbre transfer over the naive separate-then-transfer pipelines. As a result, this work confirms that the cross-stem modeling is essential for mixture-level timbre transfer as the proposed joint setting consistently exceeds an equivalent single-stem ablation.

Peiran Yan, Shengchen Li

Audio captioning is a task that generates description of audio based on content. Pre-trained models are widely used in audio captioning due to high complexity. Unless a comprehensive system is re-trained, it is hard to determine how well pre-trained models contribute to audio captioning system. To prevent the time consuming and energy consuming process of retraining, it is necessary to propose a preditor of performance for the pre-trained model in audio captioning. In this paper, a series of pre-trained models are investigated for the correlation between extracted audio features and the performance of audio captioning. A couple of predictor is proposed based on the experiment results.The result demonstrates that the kurtosis and skewness of audio features extracted may act as an indicator of the performance of audio captioning systems for pre-trained audio due to the high correlation between kurtosis and skewness of audio features and the performance of audio captioning systems.

Yuxuan Liu, Peihong Zhang, Rui Sang et al.

Music adversarial attacks have garnered significant interest in the field of Music Information Retrieval (MIR). In this paper, we present Music Adversarial Inpainting Attack (MAIA), a novel adversarial attack framework that supports both white-box and black-box attack scenarios. MAIA begins with an importance analysis to identify critical audio segments, which are then targeted for modification. Utilizing generative inpainting models, these segments are reconstructed with guidance from the output of the attacked model, ensuring subtle and effective adversarial perturbations. We evaluate MAIA on multiple MIR tasks, demonstrating high attack success rates in both white-box and black-box settings while maintaining minimal perceptual distortion. Additionally, subjective listening tests confirm the high audio fidelity of the adversarial samples. Our findings highlight vulnerabilities in current MIR systems and emphasize the need for more robust and secure models.

Peihong Zhang, Yuxuan Liu, Zhixin Li et al.

Acoustic Scene Classification (ASC) faces challenges in generalizing across recording devices, particularly when labeled data is limited. The DCASE 2024 Challenge Task 1 highlights this issue by requiring models to learn from small labeled subsets recorded on a few devices. These models need to then generalize to recordings from previously unseen devices under strict complexity constraints. While techniques such as data augmentation and the use of pre-trained models are well-established for improving model generalization, optimizing the training strategy represents a complementary yet less-explored path that introduces no additional architectural complexity or inference overhead. Among various training strategies, curriculum learning offers a promising paradigm by structuring the learning process from easier to harder examples. In this work, we propose an entropy-guided curriculum learning strategy to address the domain shift problem in data-efficient ASC. Specifically, we quantify the uncertainty of device domain predictions for each training sample by computing the Shannon entropy of the device posterior probabilities estimated by an auxiliary domain classifier. Using entropy as a proxy for domain invariance, the curriculum begins with high-entropy samples and gradually incorporates low-entropy, domain-specific ones to facilitate the learning of generalizable representations. Experimental results on multiple DCASE 2024 ASC baselines demonstrate that our strategy effectively mitigates domain shift, particularly under limited labeled data conditions. Our strategy is architecture-agnostic and introduces no additional inference cost, making it easily integrable into existing ASC baselines and offering a practical solution to domain shift.

Peihong Zhang, Zhixin Li, Rui Sang et al.

The coupling signal refers to a latent physiological signal that characterizes the transformation from cardiac electrical excitation, captured by the electrocardiogram (ECG), to mechanical contraction, recorded by the phonocardiogram (PCG). By encoding the temporal and functional interplay between electrophysiological and hemodynamic events, it serves as an intrinsic link between modalities and offers a unified representation of cardiac function, with strong potential to enhance multi-modal cardiovascular disease (CVD) detection. However, existing coupling signal estimation methods remain highly vulnerable to noise, particularly in real-world clinical and physiological settings, which undermines their robustness and limits practical value. In this study, we propose Noise-Robust Multi-Modal Coupling Signal Estimation (NMCSE), which reformulates coupling signal estimation as a distribution matching problem solved via optimal transport. By jointly aligning amplitude and timing, NMCSE avoids noise amplification and enables stable signal estimation. When integrated into a Temporal-Spatial Feature Extraction (TSFE) network, the estimated coupling signal effectively enhances multi-modal fusion for more accurate CVD detection. To evaluate robustness under real-world conditions, we design two complementary experiments targeting distinct sources of noise. The first uses the PhysioNet 2016 dataset with simulated hospital noise to assess the resilience of NMCSE to clinical interference. The second leverages the EPHNOGRAM dataset with motion-induced physiological noise to evaluate intra-state estimation stability across activity levels. Experimental results show that NMCSE consistently outperforms existing methods under both clinical and physiological noise, highlighting it as a noise-robust estimation approach that enables reliable multi-modal cardiac detection in real-world conditions.

Peihong Zhang, Zhixin Li, Yuxuan Liu et al.

Deep learning approaches for heart-sound (PCG) segmentation built on time--frequency features can be accurate but often rely on large expert-labeled datasets, limiting robustness and deployment. We present TopSeg, a topological representation-centric framework that encodes PCG dynamics with multi-scale topological features and decodes them using a lightweight temporal convolutional network (TCN) with an order- and duration-constrained inference step. To evaluate data efficiency and generalization, we train exclusively on PhysioNet 2016 dataset with subject-level subsampling and perform external validation on CirCor dataset. Under matched-capacity decoders, the topological features consistently outperform spectrogram and envelope inputs, with the largest margins at low data budgets; as a full system, TopSeg surpasses representative end-to-end baselines trained on their native inputs under the same budgets while remaining competitive at full data. Ablations at 10% training confirm that all scales contribute and that combining H_0 and H_1 yields more reliable S1/S2 localization and boundary stability. These results indicate that topology-aware representations provide a strong inductive bias for data-efficient, cross-dataset PCG segmentation, supporting practical use when labeled data are limited.

Peihong Zhang, Yuxuan Liu, Rui Sang et al.

Acoustic scene classification (ASC) suffers from device-induced domain shift, especially when labels are limited. Prior work focuses on curriculum-based training schedules that structure data presentation by ordering or reweighting training examples from easy-to-hard to facilitate learning; however, existing curricula are static, fixing the ordering or the weights before training and ignoring that example difficulty and marginal utility evolve with the learned representation. To overcome this limitation, we propose the Dynamic Dual-Signal Curriculum (DDSC), a training schedule that adapts the curriculum online by combining two signals computed each epoch: a domain-invariance signal and a learning-progress signal. A time-varying scheduler fuses these signals into per-example weights that prioritize domain-invariant examples in early epochs and progressively emphasize device-specific cases. DDSC is lightweight, architecture-agnostic, and introduces no additional inference overhead. Under the official DCASE 2024 Task~1 protocol, DDSC consistently improves cross-device performance across diverse ASC baselines and label budgets, with the largest gains on unseen-device splits.

Yiqiang Cai, Yizhou Tan, Shengchen Li et al.

Acoustic scene recordings are often collected from a diverse range of cities. Most existing acoustic scene classification (ASC) approaches focus on identifying common acoustic scene patterns across cities to enhance generalization. However, the potential acoustic differences introduced by city-specific environmental and cultural factors are overlooked. In this paper, we hypothesize that the city-specific acoustic features are beneficial for the ASC task rather than being treated as noise or bias. To this end, we propose City2Scene, a novel framework that leverages city features to improve ASC. Unlike conventional approaches that may discard or suppress city information, City2Scene transfers the city-specific knowledge from pre-trained city classification models to scene classification model using knowledge distillation. We evaluate City2Scene on three datasets of DCASE Challenge Task 1, which include both scene and city labels. Experimental results demonstrate that city features provide valuable information for classifying scenes. By distilling city-specific knowledge, City2Scene effectively improves accuracy across a variety of lightweight CNN backbones, achieving competitive performance to the top-ranked solutions of DCASE Challenge in recent years.

Xinhao Mei, Qiushi Huang, Xubo Liu et al.

Automated audio captioning aims to use natural language to describe the content of audio data. This paper presents an audio captioning system with an encoder-decoder architecture, where the decoder predicts words based on audio features extracted by the encoder. To improve the proposed system, transfer learning from either an upstream audio-related task or a large in-domain dataset is introduced to mitigate the problem induced by data scarcity. Besides, evaluation metrics are incorporated into the optimization of the model with reinforcement learning, which helps address the problem of ``exposure bias'' induced by ``teacher forcing'' training strategy and the mismatch between the evaluation metrics and the loss function. The resulting system was ranked 3rd in DCASE 2021 Task 6. Ablation studies are carried out to investigate how much each element in the proposed system can contribute to final performance. The results show that the proposed techniques significantly improve the scores of the evaluation metrics, however, reinforcement learning may impact adversely on the quality of the generated captions.

Shengchen Li, Ke Tian, Rui Wang

Heart Sound (also known as phonocardiogram (PCG)) analysis is a popular way that detects cardiovascular diseases (CVDs). Most PCG analysis uses supervised way, which demands both normal and abnormal samples. This paper proposes a method of unsupervised PCG analysis that uses beta variational auto-encoder ($β-\text{VAE}$) to model the normal PCG signals. The best performed model reaches an AUC (Area Under Curve) value of 0.91 in ROC (Receiver Operating Characteristic) test for PCG signals collected from the same source. Unlike majority of $β-\text{VAE}$s that are used as generative models, the best-performed $β-\text{VAE}$ has a $β$ value smaller than 1. Further experiments then find that the introduction of a light weighted KL divergence between distribution of latent space and normal distribution improves the performance of anomaly PCG detection based on anomaly scores resulted by reconstruction loss. The fact suggests that anomaly score based on reconstruction loss may be better than anomaly scores based on latent vectors of samples

Shengchen Li, Yinji Jing, György Fazekas

In this paper, the dataset used for the data challenge organised by Conference on Sound and Music Technology (CSMT) is introduced. The CSMT data challenge requires participants to identify whether a given piece of melody is generated by computer or is composed by human. The dataset is formed by two parts: development dataset and evaluation dataset. The development dataset contains only computer generated melodies whereas the evaluation dataset contain both computer generated melodies and human composed melodies. The aim of the dataset is to examine whether it is possible to distinguish computer generated melodies by learning the feature of generated melodies.

Yuanbo Hou, Frank K. Soong, Jian Luan et al.

Detecting singing-voice in polyphonic instrumental music is critical to music information retrieval. To train a robust vocal detector, a large dataset marked with vocal or non-vocal label at frame-level is essential. However, frame-level labeling is time-consuming and labor expensive, resulting there is little well-labeled dataset available for singing-voice detection (S-VD). Hence, we propose a data augmentation method for S-VD by transfer learning. In this study, clean speech clips with voice activity endpoints and separate instrumental music clips are artificially added together to simulate polyphonic vocals to train a vocal/non-vocal detector. Due to the different articulation and phonation between speaking and singing, the vocal detector trained with the artificial dataset does not match well with the polyphonic music which is singing vocals together with the instrumental accompaniments. To reduce this mismatch, transfer learning is used to transfer the knowledge learned from the artificial speech-plus-music training set to a small but matched polyphonic dataset, i.e., singing vocals with accompaniments. By transferring the related knowledge to make up for the lack of well-labeled training data in S-VD, the proposed data augmentation method by transfer learning can improve S-VD performance with an F-score improvement from 89.5% to 93.2%.

Yusong Wu, Shengchen Li, Chengzhu Yu et al.

Peking Opera has been the most dominant form of Chinese performing art since around 200 years ago. A Peking Opera singer usually exhibits a very strong personal style via introducing improvisation and expressiveness on stage which leads the actual rhythm and pitch contour to deviate significantly from the original music score. This inconsistency poses a great challenge in Peking Opera singing voice synthesis from a music score. In this work, we propose to deal with this issue and synthesize expressive Peking Opera singing from the music score based on the Duration Informed Attention Network (DurIAN) framework. To tackle the rhythm mismatch, Lagrange multiplier is used to find the optimal output phoneme duration sequence with the constraint of the given note duration from music score. As for the pitch contour mismatch, instead of directly inferring from music score, we adopt a pseudo music score generated from the real singing and feed it as input during training. The experiments demonstrate that with the proposed system we can synthesize Peking Opera singing voice with high-quality timbre, pitch and expressiveness.

Yusong Wu, Shengchen Li, Chengzhu Yu et al.

This paper presents a method that generates expressive singing voice of Peking opera. The synthesis of expressive opera singing usually requires pitch contours to be extracted as the training data, which relies on techniques and is not able to be manually labeled. With the Duration Informed Attention Network (DurIAN), this paper makes use of musical note instead of pitch contours for expressive opera singing synthesis. The proposed method enables human annotation being combined with automatic extracted features to be used as training data thus the proposed method gives extra flexibility in data collection for Peking opera singing synthesis. Comparing with the expressive singing voice of Peking opera synthesised by pitch contour based system, the proposed musical note based system produces comparable singing voice in Peking opera with expressiveness in various aspects.

Yuanbo Hou, Qiuqiang Kong, Shengchen Li et al.

Sound event detection (SED) methods typically rely on either strongly labelled data or weakly labelled data. As an alternative, sequentially labelled data (SLD) was proposed. In SLD, the events and the order of events in audio clips are known, without knowing the occurrence time of events. This paper proposes a connectionist temporal classification (CTC) based SED system that uses SLD instead of strongly labelled data, with a novel unsupervised clustering stage. Experiments on 41 classes of sound events show that the proposed two-stage method trained on SLD achieves performance comparable to the previous state-of-the-art SED system trained on strongly labelled data, and is far better than another state-of-the-art SED system trained on weakly labelled data, which indicates the effectiveness of the proposed two-stage method trained on SLD without any onset/offset time of sound events.

Yuanbo Hou, Qiuqiang Kong, Jun Wang et al.

Audio tagging aims to detect the types of sound events occurring in an audio recording. To tag the polyphonic audio recordings, we propose to use Connectionist Temporal Classification (CTC) loss function on the top of Convolutional Recurrent Neural Network (CRNN) with learnable Gated Linear Units (GLU-CTC), based on a new type of audio label data: Sequentially Labelled Data (SLD). In GLU-CTC, CTC objective function maps the frame-level probability of labels to clip-level probability of labels. To compare the mapping ability of GLU-CTC for sound events, we train a CRNN with GLU based on Global Max Pooling (GLU-GMP) and a CRNN with GLU based on Global Average Pooling (GLU-GAP). And we also compare the proposed GLU-CTC system with the baseline system, which is a CRNN trained using CTC loss function without GLU. The experiments show that the GLU-CTC achieves an Area Under Curve (AUC) score of 0.882 in audio tagging, outperforming the GLU-GMP of 0.803, GLU-GAP of 0.766 and baseline system of 0.837. That means based on the same CRNN model with GLU, the performance of CTC mapping is better than the GMP and GAP mapping. Given both based on the CTC mapping, the CRNN with GLU outperforms the CRNN without GLU.

Yuanbo Hou, Qiuqiang Kong, Shengchen Li

Audio tagging aims to predict one or several labels in an audio clip. Many previous works use weakly labelled data (WLD) for audio tagging, where only presence or absence of sound events is known, but the order of sound events is unknown. To use the order information of sound events, we propose sequential labelled data (SLD), where both the presence or absence and the order information of sound events are known. To utilize SLD in audio tagging, we propose a Convolutional Recurrent Neural Network followed by a Connectionist Temporal Classification (CRNN-CTC) objective function to map from an audio clip spectrogram to SLD. Experiments show that CRNN-CTC obtains an Area Under Curve (AUC) score of 0.986 in audio tagging, outperforming the baseline CRNN of 0.908 and 0.815 with Max Pooling and Average Pooling, respectively. In addition, we show CRNN-CTC has the ability to predict the order of sound events in an audio clip.