Event Specific Attention for Polyphonic Sound Event Detection

Sundar, Harshavardhan; Sun, Ming; Wang, Chao

doi:10.21437/interspeech.2021-684

Cited by 11 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To investigate the effectiveness of the proposed MGA-Net, we compare it with the state-of-the-art methods [13,27]. As shown in Table 1, the MGA-Net achieves 53.27%, and 56.96% EB-F1 score, 0.709 and 0.739 PSDS score for the validation and public set, respectively, significantly outperforming the compared methods.…”

Section: Resultsmentioning

confidence: 99%

“…The global context modeling is built upon the multi-head selfattention mechanism [25]. Considering the sequential position of input features, we introduce relative positional encoding (RPE) [26] which has been shown effective in SED task [27] to encode position information of inter-frames. The length of attention weights is that of the entire time making the feature representation more global but coarser.…”

Section: Global Context Modelingmentioning

confidence: 99%

See 1 more Smart Citation

A Multi-grained based Attention Network for Semi-supervised Sound Event Detection

Hu,

Zhu,

et al. 2022

Preprint

View full text Add to dashboard Cite

Sound event detection (SED) is an interesting but challenging task due to the scarcity of data and diverse sound events in real life. This paper presents a multi-grained based attention network (MGA-Net) for semi-supervised sound event detection. To obtain the feature representations related to sound events, a residual hybrid convolution (RH-Conv) block is designed to boost the vanilla convolution's ability to extract the time-frequency features. Moreover, a multi-grained attention (MGA) module is designed to learn temporal resolution features from coarse-level to fine-level. With the MGA module, the network could capture the characteristics of target events with short-or long-duration, resulting in more accurately determining the onset and offset of sound events. Furthermore, to effectively boost the performance of the Mean Teacher (MT) method, a spatial shift (SS) module as a data perturbation mechanism is introduced to increase the diversity of data. Experimental results show that the MGA-Net outperforms the published state-of-the-art competitors, achieving 53.27% and 56.96 % event-based macro F1 (EB-F1) score, 0.709 and 0.739 polyphonic sound detection score (PSDS) on the validation and public set respectively.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Global Context Modelingmentioning

confidence: 99%

A Multi-grained based Attention Network for Semi-supervised Sound Event Detection

Hu,

Zhu,

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Some scholars indirectly quantify investor attention using metrics like news headlines (Barber and Odean 2008) and advertising (Hsu and Chen 2019;Yang et al 2021) to examine investor attention. Meanwhile, others identify changes in attention triggered by specific stock market events (Sundar et al 2021), employing event study methodology (Li and Wu 2024). These studies face a significant challenge, as indirect measurements are effective only when investors genuinely notice and read the information.…”

Section: Limited Investor Attentionmentioning

confidence: 99%

Impacts of Investor Attention and Accounting Information Comparability on Stock Returns: Empirical Evidence from Chinese Listed Companies

Zhao,

Naktnasukanjn,

Dawod

et al. 2024

IJFS

View full text Add to dashboard Cite

The efficient capital markets hypothesis (EMH) posits that security prices incorporate all available information in capital markets. Nevertheless, real stock markets often exhibit speculative behavior due to information asymmetry and the limited rationality of investors. This paper employs Thank you for your feedback. I've reviewed the changes, and the adjustment made to revise the year from 2019 to 2023 according to the reference information is accurate. statistical analysis, a multiple regression approach, and robustness tests to investigate the impact of investor attention and accounting information comparability on stock returns. We collected monthly data from all Chinese A-share stocks listed on the main board of the Shanghai Stock Exchange for the period 2017–2021. Our findings reveal a significant positive correlation between current investor attention and current monthly stock returns and a significant negative correlation between lagged investor attention and current monthly stock returns. Moreover, accounting information comparability serves as a substantial moderator, amplifying the positive effect of current investor attention on current stock returns and mitigating the negative impact of lagged investor attention. We investigate the indicator of accounting information comparability from the perspective of investor attention. Significantly, we use accounting information comparability as a moderating variable for the first time to assess its influence on stock returns. Our results demonstrate that accounting information comparability significantly contributes to mitigating excessive share price declines and stimulating share price increases. This discovery also acts as an internal driver for listed companies to proactively improve accounting information comparability.

show abstract

“…These neural networks are usually trained over audio features such as the Short-Time Fourier Transform (STFT), mel-spectrograms, or mel-frequency cepstral coefficients (MFCC) [16], in which a bank of filters is applied to short segments of the input audio signal, obtaining a time-frequency representation of the audio. Some recent approaches have also employed attention-based networks [17][18][19] like Conformers [20], which were originally proposed for Speech Recognition.…”

Section: Introductionmentioning

confidence: 99%

An Analysis of Sound Event Detection under Acoustic Degradation Using Multi-Resolution Systems

2021

View full text Add to dashboard Cite

The Sound Event Detection task aims to determine the temporal locations of acoustic events in audio clips. In recent years, the relevance of this field is rising due to the introduction of datasets such as Google AudioSet or DESED (Domestic Environment Sound Event Detection) and competitive evaluations like the DCASE Challenge (Detection and Classification of Acoustic Scenes and Events). In this paper, we analyze the performance of Sound Event Detection systems under diverse artificial acoustic conditions such as high- or low-pass filtering and clipping or dynamic range compression, as well as under an scenario of high overlap between events. For this purpose, the audio was obtained from the Evaluation subset of the DESED dataset, whereas the systems were trained in the context of the DCASE Challenge 2020 Task 4. Our systems are based upon the challenge baseline, which consists of a Convolutional-Recurrent Neural Network trained using the Mean Teacher method, and they employ a multiresolution approach which is able to improve the Sound Event Detection performance through the use of several resolutions during the extraction of Mel-spectrogram features. We provide insights on the benefits of this multiresolution approach in different acoustic settings, and compare the performance of the single-resolution systems in the aforementioned scenarios when using different resolutions. Furthermore, we complement the analysis of the performance in the high-overlap scenario by assessing the degree of overlap of each event category in sound event detection datasets.

show abstract

Event Specific Attention for Polyphonic Sound Event Detection

Cited by 11 publications

References 0 publications

A Multi-grained based Attention Network for Semi-supervised Sound Event Detection

A Multi-grained based Attention Network for Semi-supervised Sound Event Detection

Impacts of Investor Attention and Accounting Information Comparability on Stock Returns: Empirical Evidence from Chinese Listed Companies

An Analysis of Sound Event Detection under Acoustic Degradation Using Multi-Resolution Systems

Contact Info

Product

Resources

About