Annu Int Conf IEEE Eng Med Biol Soc. 2021 Nov;2021:2226-2229. doi: 10.1109/EMBC46164.2021.9630699.
Phonological categories in articulated speech are defined based on the place and manner of articulation. In this work, we investigate whether the phonological categories of the prompts imagined during speech imagery lead to differences in phase synchronization in various cortical regions that can be discriminated from the EEG captured during the imagination. Nasal and bilabial consonant are the two phonological categories considered due to their differences in both place and manner of articulation. Mean phase coherence (MPC) is used for measuring the phase synchronization and shallow neural network (NN) is used as the classifier. As a benchmark, we have also designed another NN based on statistical parameters extracted from imagined speech EEG. The NN trained on MPC values in the beta band gives classification results superior to NN trained on alpha band MPC values, gamma band MPC values and statistical parameters extracted from the EEG.Clinical relevance: Brain-computer interface (BCI) is a promising tool for aiding differently-abled people and for neurorehabilitation. One of the challenges in designing speech imagery based BCI is the identification of speech prompts that can lead to distinct neural activations. We have shown that nasal and blilabial consonants lead to dissimilar activations. Hence prompts orthogonal in these phonological categories are good choices as speech imagery prompts.
在有声音语言中，语音类别是基于发音部位和发音方式来定义的。在这项工作中，我们研究了在语音想象过程中想象的提示的语音类别是否会导致各个皮质区域的相位同步产生差异，从而可以从想象过程中捕获的 EEG 中区分出来。由于发音部位和发音方式的不同，我们考虑了鼻音和双唇音这两个语音类别。平均相位相干性（MPC）用于测量相位同步，浅层神经网络（NN）用作分类器。作为基准，我们还基于想象中的语音 EEG 提取的统计参数设计了另一个 NN。在 beta 波段的 MPC 值上训练的 NN 给出的分类结果优于在 alpha 波段 MPC 值、gamma 波段 MPC 值和从 EEG 中提取的统计参数上训练的 NN。临床相关性：脑机接口（BCI）是一种有前途的辅助残疾人和神经康复的工具。基于语音想象设计 BCI 的挑战之一是确定可以导致不同神经激活的语音提示。我们已经表明，鼻音和双唇音会导致不同的激活。因此，在这些语音类别中正交的提示是作为语音想象提示的不错选择。
