A Dataset of Dynamic Reverberant Sound Scenes with Directional Interferers for Sound Event Localization and Detection

Politis, Archontis; Adavanne, Sharath; Krause, Daniel; Deleforge, Antoine; Srivastava, Prerak; Virtanen, Tuomas

doi:10.48550/arxiv.2106.06999

Cited by 10 publications

(27 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That enables the model to solve a SELD task with a single objective such as the mean squared error (MSE) using a single network. As the single-ACCDOA format outperforms the two-branch format in the joint SELD metrics with a simpler architecture [11], ACCDOA was adopted to the DCASE 2021 Task 3 baseline [12]. A convolutional recurrent neural-network (CRNN), consisting of convolution blocks, gated recurrent unit (GRU) layers, and a fully-connected layer, is used to estimate ACCDOA vectors in [11].…”

Section: Related Workmentioning

confidence: 99%

“…We evaluated the multi-ACCDOA format trained with three PITs in Sec. 3, the single-ACCDOA format, and the track-wise format using TAU Spatial Sound Events 2021 [12]. The proposed method is also compared with state-of-the-art SELD methods.…”

Section: Experimental Evaluationmentioning

confidence: 99%

“…1 shows the overview of the proposed method. In experimental evaluations using the DCASE 2021 Task 3 dataset [12], the result shows the multi-ACCDOA method with the class-wise ADPIT detected over-lapping events from the same class without any degradation in the other cases. Also, the proposed method performed comparably to state-of-the-art methods with fewer parameters.…”

Section: Introductionmentioning

confidence: 99%

“…NN-based SELD methods can be categorized into two output formats. The first is the class-wise output format, in which the model predicts activities of all event classes and corresponding locations [7][8][9][10][11][12][13]. Adavanne et al proposed SELDnet, which detects sound events and estimates the corresponding DOAs using two branches: an sound event detection (SED) branch and a DOA branch [7,8].…”

Section: Introductionmentioning

confidence: 99%

“…Adavanne et al proposed SELDnet, which detects sound events and estimates the corresponding DOAs using two branches: an sound event detection (SED) branch and a DOA branch [7,8]. Activity-coupled Cartesian DOA (ACCDOA) assigns an event activity to the length of a corresponding Cartesian DOA vector [11,12], which enables a SELD task to be solved without branching. In the evaluations on the SELD task for DCASE 2020 Task 3, the single-ACCDOA format outperformed the two-branch format with fewer parameters [11].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Multi-ACCDOA: Localizing and Detecting Overlapping Sounds from the Same Class with Auxiliary Duplicating Permutation Invariant Training

Shimada¹,

Koyama²,

Takahashi³

et al. 2021

Preprint

View full text Add to dashboard Cite

Sound event localization and detection (SELD) involves identifying the direction-of-arrival (DOA) and the event class. The SELD methods with a class-wise output format make the model predict activities of all sound event classes and corresponding locations. The class-wise methods can output activity-coupled Cartesian DOA (AC-CDOA) vectors, which enable us to solve a SELD task with a single target using a single network. However, there is still a challenge in detecting the same event class from multiple locations. To overcome this problem while maintaining the advantages of the class-wise format, we extended ACCDOA to a multi one and proposed auxiliary duplicating permutation invariant training (ADPIT). The multi-ACCDOA format (a class-and track-wise output format) enables the model to solve the cases with overlaps from the same class. The class-wise ADPIT scheme enables each track of the multi-ACCDOA format to learn with the same target as the single-ACCDOA format. In evaluations with the DCASE 2021 Task 3 dataset, the model trained with the multi-ACCDOA format and with the class-wise AD-PIT detects overlapping events from the same class while maintaining its performance in the other cases. Also, the proposed method performed comparably to state-of-the-art SELD methods with fewer parameters.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Experimental Evaluationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Multi-ACCDOA: Localizing and Detecting Overlapping Sounds from the Same Class with Auxiliary Duplicating Permutation Invariant Training

Shimada¹,

Koyama²,

Takahashi³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

Multi-ACCDOA: Localizing and Detecting Overlapping Sounds from the Same Class with Auxiliary Duplicating Permutation Invariant Training

Shimada¹,

Koyama²,

Takahashi³

et al. 2022

Preprint

View full text Add to dashboard Cite

Sound event localization and detection (SELD) involves identifying the direction-of-arrival (DOA) and the event class. The SELD methods with a class-wise output format make the model predict activities of all sound event classes and corresponding locations. The class-wise methods can output activity-coupled Cartesian DOA (AC-CDOA) vectors, which enable us to solve a SELD task with a single target using a single network. However, there is still a challenge in detecting the same event class from multiple locations. To overcome this problem while maintaining the advantages of the class-wise format, we extended ACCDOA to a multi one and proposed auxiliary duplicating permutation invariant training (ADPIT). The multi-ACCDOA format (a class- and track-wise output format) enablesthe model to solve the cases with overlaps from the same class. The class-wise ADPIT scheme enables each track of the multi-ACCDOA format to learn with the same target as the single-ACCDOA format. In evaluations with the DCASE 2021 Task 3 dataset, the model trained with the multi-ACCDOA format and with the class-wise ADPIT detects overlapping events from the same class while maintain-ing its performance in the other cases. Also, the proposed method performed comparably to state-of-the-art SELD methods with fewer parameters.

show abstract

A Method Based on Dual Cross-Modal Attention and Parameter Sharing for Polyphonic Sound Event Localization and Detection

et al. 2022

View full text Add to dashboard Cite

Sound event localization and detection (SELD) is a joint task that unifies sound event detection (SED) and direction-of-arrival estimation (DOAE). The task has become such a popular topic that it was introduced into the challenge on Detection and Classification of Acoustic Scenes and Events (DCASE) Task3 in 2019. In this paper, we propose a method based on dual cross-modal attention (DCMA) and parameter sharing to simultaneously detect and localize sound events. In particular, the DCMA-based decoder commonly used for multiple predictions efficiently learns the associations between SED and DOAE features by exchanging SED and DOAE information in the process of attention, in addition to the encoder with parameter sharing. Furthermore, acoustic features that have not been usually used in the SELD task are additionally adopted to improve the performance, and data augmentation techniques of the mixup to simulate polyphonic events and channel rotation for spatial augmentation are conducted for this task. Experimental results demonstrate that our efficient model using one common decoder block based on the DCMA to predict multiple events in the track-wise output format is effective for the SELD task with up to three overlapping events.

show abstract

A Dataset of Dynamic Reverberant Sound Scenes with Directional Interferers for Sound Event Localization and Detection

Cited by 10 publications

References 0 publications

Multi-ACCDOA: Localizing and Detecting Overlapping Sounds from the Same Class with Auxiliary Duplicating Permutation Invariant Training

Multi-ACCDOA: Localizing and Detecting Overlapping Sounds from the Same Class with Auxiliary Duplicating Permutation Invariant Training

Multi-ACCDOA: Localizing and Detecting Overlapping Sounds from the Same Class with Auxiliary Duplicating Permutation Invariant Training

A Method Based on Dual Cross-Modal Attention and Parameter Sharing for Polyphonic Sound Event Localization and Detection

Contact Info

Product

Resources

About