Intelligent Speech Recognition Technology in Comparative Research of Foreign Language Speech Based on Whisper Model Recurrent Neural Network

In recent years, the widespread of automatic speech recognition technology in applications like emergency response centers, virtual assistants and search engines had highlighted need for system that effectively handle diverse ways in which people communicate. Traditional approaches for speech recognition in foreign language speech had faced several challenges which include assent and pronunciation, limited training data. Therefore, this research proposes Whisper Model Recurrent Neural Network namely (Whisper model RNN) for recognizing intelligent speech in foreign language. Initially, the data is collected from Texas Instruments Massachusetts Institute of Technology (TIMIT) dataset. Then, the collected data is further preprocessed by using Spectral Subtraction then, these features are extracted by using Mel-Frequency Cepstral Coefficients (MFCC) with Discrete Wavelet Transform (DWT) which efficiently captures the spectral and temporal characteristics of the speech signal, resulting in a robust and compact feature representation. Finally, recognition of speech is done by proposed Whisper model RNN which learns complex patterns and relationships in the speech data by achieving high accuracy and robustness in speech recognition. The proposed model achieved better accuracy with consistency rate of 93.87% and accuracy with adjacent consistency rate of 99.91% when compared with existing Mobile sensor networks.