A Learnable Spatial Mapping for Decoding the Directional Focus of Auditory Attention Using EEG

Zhang, Yuanming; Ruan, Haoxin; Yuan, Zhengrong; Du, Haoliang; Xia, Guoqing; Lü, Jing

doi:10.1109/icassp49357.2023.10096819

Cited by 3 publications

(5 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The subject-independent evaluation demonstrates that the CNN baseline model achieves an accuracy of 80.4%, demonstrating its ability to extract general patterns for directional focus decoding across different listeners. The integration of the LSM module into the CNN model results in a higher accuracy (85.7% on the subject-independent dataset), which is consistent with our prior findings in binary decoding [21]. AEP-CRN has demonstrated its ability to attain state-of-the-art performance in binary decoding [17].…”

Section: A the Feasibility Of Multidirectional Decodingsupporting

confidence: 88%

“…Table IV demonstrates that the decision window length, which represents the time duration of EEG signals, strongly influences decoding accuracy. Previous research on binary decoding has shown that both CNN and LSM-CNN models exhibit improved performance with longer decision window lengths [21]. Surprisingly, in multidirectional decoding, all models achieve the highest accuracy with a decision window length of 1 second.…”

Section: The Impact Of Decision Window Lengthmentioning

confidence: 79%

“…A convolutional recurrent network (CRN) is then applied to extract patterns in the 3D representation (2 spatial dimensions and 1 temporal dimension) of EEG. Recently, we have proposed a learnable spatial mapping (LSM) mechanism [21] to transform the EEG channels into a 2D form, which has been shown to be able to boost the performance of the CNN-based model. However, its benefit over the rule-based AEP method and the flexibility of merging into other models are yet to be investigated.…”

Section: Electroencephalogram-based Multiclass Auditory Attention Dec...mentioning

confidence: 99%

“…We have constructed a new dataset, the NJU auditory attention decoding dataset, containing multiple alternative directions of attended speakers [21], [26]. Our previous research focused on binary directional focus decoding [21], and this paper will further explore the feasibility of multiclass directional focus decoding.…”

Section: A Experiments Setupmentioning

confidence: 99%

“…Subsequently, a manual inspection is conducted to identify and eliminate additional abnormal segments within the EEG signal. Before being fed into the DNN model, the EEG signals are split into four subbands through four bandpass filters: the delta band (1-4 Hz), theta band (4-8 Hz), alpha band (8-13 Hz), and beta band (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) Hz). The dataset is available at https://ieee-dataport.org/documents/nju-auditory-attention-dec oding-dataset.…”

Section: B Eeg Signal Preprocessingmentioning

confidence: 99%

See 4 more Smart Citations

Electroencephalogram-based Multiclass Auditory Attention Decoding of Attended Speaker Direction

Zhang,

Lu,

Lin

et al. 2023

Preprint

View full text Add to dashboard Cite

<p>Decoding the directional focus of an attended speaker from listeners’ electroencephalogram signals is an important part of a practical brain-computer interface device aimed at improving the quality of life for individuals with hearing impairment. Existing works focus on binary directional focus decoding, i.e., determining whether the attended speaker is on the left or right side of the listener. However, the information brought by the binary decoding to the subsequent speech processing algorithm is limited in practical applications, and more precise decoding of the exact direction of the attended speaker is desired. In this paper, we first present a new dataset with 15 alternative speaker directions, and then demonstrate the feasibility of multiclass directional focus decoding of attended speakers by applying our recently proposed learnable spatial mapping (LSM) module, the benefit of which has already been proven in binary decoding scenarios. Apart from combining the LSM module with the convolutional neural network (CNN), we further validate its benefit by combining it with the spectro-spatial-temporal convolutional recurrent network (CRN), a recently proposed state-of-the-art model for binary directional focus decoding. The proposed LSM-CNN and LSM-CRN models achieve a noteworthy decoding accuracy of 85.7% and 87.5%, respectively, in the presented subject-independent dataset with a decision window length of 1 second. Comprehensive experiments not only substantiate the advantages attributed to the LSM module, but also verify the influence of the decision window length, the distraction caused by interfering speakers, and the contribution of different subband of EEG signals.</p>

show abstract

Section: A the Feasibility Of Multidirectional Decodingsupporting

confidence: 88%

Section: The Impact Of Decision Window Lengthmentioning

confidence: 79%

Section: Electroencephalogram-based Multiclass Auditory Attention Dec...mentioning

confidence: 99%

Section: A Experiments Setupmentioning

confidence: 99%