GTSO: Gradient tangent search optimization enabled voice transformer with speech intelligibility for aphasia

Abstract

A frequent neurological condition known as aphasia is brought on by injury to language-related brain regions as well as possibly other regions of the brain involved in executive, memory, and attention functions. Due to a lack of speech-language pathologists and the vast expense of treatment, traditional therapy is difficult for aphasia-affected people to access. In this research work, speech intelligibility for aphasia is done by the proposed Gradient Tangent Search Optimization (GTSO) algorithm-enabled voice transformer. Here, the median filter is used for pre-processing the signal to reduce noise. The pre-processed voice signal is allowed for feature extraction and voice enhancement stages. Moreover, nonlinear spectral subtraction is used for voice enhancement and voice transformer is used for voice recognition. Also, the voice transformer is trained by GTSO, which is devised by hybridizing Gradient Descent (GD) Optimization and Tangent Search Algorithm (TSA). Then, the output obtained is fed to the language and pronunciation model for recognizing speech, and at last, the speech recognized is converted to text. Furthermore, the GTSO- enabled voice transformer is analyzed for its performance by three metrics, namely recognition accuracy, Positive Predictive Value (PPV), and Negative Predictive Value (NPV), with superior values of 0.919, 0.919, and 0.915.