With the advancement of engineering solutions in the medical domain, the patient’s life can become comfortable. This work recognizes the silent speech of three words. The decoding of silent speech can be useful for patients who are in a locked-in syndrome state. Moreover, it is also applicable to entertainment, cognitive biometrics, and brain-computer interfacing. Brain waves of these imagined words in the delta, theta, alpha, beta, gamma, and high gamma frequency bands are analysed. Covariance based connectivity features are extracted in each frequency band. The principal features which represent more than 95% of the variance are selected as a subset of the covariance connectivity matrix. This sub-set is tested on five classifiers. The maximum accuracy achieved is 76.4% in the theta band. Also, theta and high gamma band contain maximum information about imagined speech with average accuracies of 68.32% and 62.09% respectively.

中文翻译：

---

使用基于协方差的子集选择方法将想象的思想转换为语音

随着医学领域工程解决方案的发展，患者的生活将变得舒适。这项工作承认了三个单词的无声讲话。沉默语音的解码对于处于锁定综合征状态的患者可能很有用。此外，它还适用于娱乐，认知生物识别技术和脑机接口。分析了三角形，θ，α，β，伽玛和高伽玛频段中这些虚构单词的脑电波。在每个频带中提取基于协方差的连通性特征。选择代表超过95％的方差的主要特征作为协方差连通性矩阵的子集。该子集在五个分类器上进行了测试。在θ带中达到的最高准确度为76.4％。还，