An Interpretable Deep Learning Model for Speech Activity Detection Using Electrocorticographic Signals

An Interpretable Deep Learning Model for Speech Activity Detection Using Electrocorticographic Signals
复制标题

使用皮层电信号进行语音活动检测的可解释深度学习模型

DOI:
10.1109/tnsre.2022.3207624
复制
发表时间:
2022
影响因子:
4.9
通讯作者:
Krusienski, Dean J.
Krusienski, Dean J.
中科院分区:
工程技术2区
文献类型:
--
作者:
Stuart, Morgan;Lesaja, Srdjan;Shih, Jerry J.;Schultz, Tanja;Manic, Milos;Krusienski, Dean J.

文献摘要

参考文献

相似文献

许多最先进的神经语音解码和合成解决方案将深度学习纳入处理管道。这些模型通常是不透明的,并且可能需要大量的计算资源来进行训练和执行。提出了一种深度学习架构,该架构学习直接从数据中捕获任务相关频谱特征的输入带通滤波器。将这种可解释的特征提取扩展到模型中进一步实现了创建端到端架构的目标,该架构能够自动进行特定于主题的参数调整,同时产生可解释的结果。该模型使用在语音任务期间收集的颅内脑数据来实现。使用原始的,未经处理的时间样本,该模型检测语音的存在,在每一个时间样本的因果关系的方式,适合在线应用。模型性能与需要大量信号预处理的现有方法相当或上级,并且发现学习的频带收敛到由先前研究支持的范围。
Numerous state-of-the-art solutions for neural speech decoding and synthesis incorporate deep learning into the processing pipeline. These models are typically opaque and can require significant computational resources for training and execution. A deep learning architecture is presented that learns input bandpass filters that capture task-relevant spectral features directly from data. Incorporating such explainable feature extraction into the model furthers the goal of creating end-to-end architectures that enable automated subject-specific parameter tuning while yielding an interpretable result. The model is implemented using intracranial brain data collected during a speech task. Using raw, unprocessed timesamples, the model detects the presence of speech at every timesample in a causal manner, suitable for online application. Model performance is comparable or superior to existing approaches that require substantial signal preprocessing and the learned frequency bands were found to converge to ranges that are supported by previous studies.
DOI: 10.1038/srep25803
发表时间: 2016-05-11
期刊: Scientific reports
影响因子: 4.6
作者:
Martin S;Brunner P;Iturrate I;Millán Jdel R;Schalk G;Knight RT;Pasley BN
通讯作者: Pasley BN
使用具有线性相位卷积的神经网络进行癫痫尖峰检测
DOI: 10.1109/jbhi.2021.3102247
发表时间: 2021
期刊: IEEE J Biomed Health Inform.
影响因子: --
作者:
Fukumori K;Yoshida N;Sugano H;Nakajima M;Tanaka T
通讯作者: Tanaka T
DOI: 10.1109/tbme.2014.2298897
发表时间: 2014-04
期刊: IEEE transactions on bio-medical engineering
影响因子: --
作者:
Kanas VG;Mporas I;Benz HL;Sgarbas KN;Bezerianos A;Crone NE
通讯作者: Crone NE
并发任务范式下多语言者言语产生过程中的偏侧效应:习得年龄假说的另一个检验
DOI: --
发表时间: 2003
期刊:
影响因子: --
作者:
M. Patkowski
通讯作者: M. Patkowski
DOI: 10.1088/1741-2552/abda0b
发表时间: 2021-04-01
影响因子: 4
作者:
Peterson, Steven M.;Steine-Hanson, Zoe;Brunton, Bingni W.
通讯作者: Brunton, Bingni W.