Sound Events Localization and Detection Using Bio-inspired Gammatone Filters and Temporal Convolutional Neural Networks

Rosero, Karen; Grijalva, Felipe; Masiero, Bruno

doi:10.36227/techrxiv.21155257

Cited by 2 publications

(3 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first dataset comprises four couples of recorded signals, including periodic noisy clapping sounds at positions (0,0), (0,1.5), (1,1.5), and (1,2). Two sensors are located on the ground at positions (0,0) and (2,3). The mean of the background noise is 0.42 (W) with a standard deviation of 5 (kW), and the signal-to-noise ratio ranges between 5 to 8 dB.…”

Section: A Datasetmentioning

confidence: 99%

“…[2] exploits the echo state SNN capability synergized with CNN classification methods, resulting in enhanced accuracy. Furthermore, [3] employs Convolutional Recurrent Neural Network (CRNN) methods incorporating Gammatone filtering and frequencybased approaches, yielding promising results. These multifaceted methodologies showcase the evolution of SSL tech-niques, embracing diverse technologies and demonstrating promising outcomes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Dynamic-Structured Reservoir Spiking Neural Network in Sound Localization

Roozbehi,

Narayanan,

Mohaghegh

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Sound source localization is a critical problem in various fields, including communication, security, and entertainment. Binaural cues are a natural technique used by mammalian ears for efficient sound source localization. Spiking neural networks (SNNs) have emerged as a promising tool for implementing binaural sound source localization approaches. However, optimizing the topology and size of SNNs is crucial to reduce computational costs while maintaining accuracy. This paper proposes a realtime structure of a reservoir SNN (rSNN) called Adaptive-Resonance-Theory-based rSNN (ART-rSNN) for localizing sound sources in the time domain by integrating an energy-based localization method. The dataset used in this work is recorded by two different omnidirectional microphones from a real environment. The dataset includes various sound events such as speech, music, and environmental sounds. The proposed ART-rSNN architecture can dynamically adjust the location of its neurons to amplify estimated energy near the sound source, resulting in higher localization accuracy. Our proposed method outperforms several conventional and state of the art algorithms in terms of accuracy and is able to detect the front and back direction of azimuth angle. This work demonstrates the potential of dynamic neuron arrangements in SNNs for improving sound source localization in practical applications

show abstract

Section: A Datasetmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Dynamic-Structured Reservoir Spiking Neural Network in Sound Localization

Roozbehi,

Narayanan,

Mohaghegh

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…It was proved crucial for addressing overlapping sound events. Rosero et al devised a Gammatone-based Sound Events Localization and Detection (G-SELD) system [13], which seeks to enhance the SSLD performance by employing the bio-inspired gammatone auditory filters for the acoustic feature extraction.…”

Section: Related Workmentioning

confidence: 99%

GLFER-Net: a polyphonic sound source localization and detection network based on global-local feature extraction and recalibration

Ma,

Hu,

et al. 2024

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

Polyphonic sound source localization and detection (SSLD) task aims to recognize the categories of sound events, identify their onset and offset times, and detect their corresponding direction-of-arrival (DOA), where polyphonic refers to the occurrence of multiple overlapping sound sources in a segment. However, vanilla SSLD methods based on convolutional recurrent neural network (CRNN) suffer from insufficient feature extraction. The convolutions with kernel of single scale in CRNN fail to adequately extract multi-scale features of sound events, which have diverse time-frequency characteristics. It results in that the extracted features lack fine-grained information helpful for the localization of sound sources. In response to these challenges, we propose a polyphonic SSLD network based on global-local feature extraction and recalibration (GLFER-Net), where the global-local feature (GLF) extractor is designed to extract the multi-scale global features through an omni-directional dynamic convolution (ODConv) layer and multi-scale feature extraction (MSFE) module. The local feature extraction (LFE) unit is designed for capturing detailed information. Besides, we design a feature recalibration (FR) module to emphasize the crucial features along multiple dimensions. On the open datasets of Task3 in DCASE 2021 and 2022 Challenges, we compared our proposed GLFER-Net with six and four SSLD methods, respectively. The results show that the GLFER-Net achieves competitive performance. The modules we designed are verified to be effective through a series of ablation experiments and visualization analyses.

show abstract

Sound Events Localization and Detection Using Bio-inspired Gammatone Filters and Temporal Convolutional Neural Networks

Cited by 2 publications

References 4 publications

Dynamic-Structured Reservoir Spiking Neural Network in Sound Localization

Dynamic-Structured Reservoir Spiking Neural Network in Sound Localization

GLFER-Net: a polyphonic sound source localization and detection network based on global-local feature extraction and recalibration

Contact Info

Product

Resources

About