Article Preview
Top1. Introduction
Speech is the basic mode of communication and interaction among the community utilizing spoken language. Human beings with speech disorders face several issues in daily life, which leads to emotional instability, which further isolates them from society. The factors like fluidity, intensity, and pitch are used to measure the voice disorder. The voice production is decided by the laryngeal strength of muscle, supraglottic resonator cavities like nasal cavity, pharyngeal and oral, and airflow. In addition, the respiratory system supplies the airflow; hence the vibration of the vocal cords produces the voice (Siyuan, et al., 2019). Psychogenic, functional, and organic issues are the different reasons for any voice disorder (Lee, et al., 2021). For example, some humans cannot normally communicate like other people due to the locked-in syndrome. These persons cannot do any motor actions like movements of hands and feet (Fabien & Roy, 2019), and they cannot speak. Still, they can function and recognize the nearby environment by their cognitive system (Dipti & Dhage, 2020).
The cognitive function and the brain of abnormal hearing humans work properly. However, due to technology development, smart applications are devised without human intervention to control and manage abnormal hearing humans using sensor techniques. An example of this is the Internet of Things (IoT). The actuators and sensors are implanted in the physical object, and the communication is performed using either a wired or wireless network. Several existing methodologies are developed to express the feelings and communication of the hearing impaired humans using the acquisition and interpretation of brain signals by utilizing several sensors. The above mentioned technique of acquiring the brain signals figuring speech is termed imaginary speech (Dash, et al., 2019; Kumar, et al., 2018). In past years, imaginary speech has been emerging research using the brain's signals without the audible voice requirement (Kumar & Singh, 2019). It consists of phonemes or words without any articulate movement or voice. Imaginary speech recognition is widely used in several applications for motor disabilities and hearing (Xiaorong, et al., 2021).
Several conventional techniques used for the imaginary speech classification are employed based on the Electroencephalogram (EEG) method (Li, et al., 2021). The EEG has the advantage of better temporal resolution and is cost-effective compared to other existing techniques (Sumi, et al., 2019). The imagined speech recognition using EEG signals provides better results with increased efficiency (Sgro, et al., 2019). First, imagined speech recognition is employed by extracting features from EEG signals (Magdiel & Gómez-Gil, 2021). Then classification is employed for recognition of imagined speech. The commonly used technique in (Xinyu, et al., 2020) is pre-processing using bandpass filtering and feature extraction using the spatial filter. Finally, classification is employed by using the SVM. The Common spatial pattern (CSP) filter is used for the feature extraction, but an improved form of CSP is also devised in several techniques. The classification of EEG signals is employed in different forms: matrix and tensor classifiers, deep learning, machine learning (Alzubi, et al., 2018), adaptive classifiers, and transfer learning technique (Bakhshali, et al., 2020). The optimization based classification enhances the accuracy of the system (Nayyar, et al., 2018). Most optimization approaches are inspired by swarm intelligence due to their flexibility and robustness (Nayyar, et al., 2018). Besides, it is helpful to solve complex optimization issues (Nayyar & Nguyen, 2018) . However, better accurate techniques are developed by improving the designs with the enhanced systems (Sharon & Murthy, 2020).