Improving the Performance of Speech Recognition feature selection using Northern Goshawk Optimization

Abstract

With the advancement of digital signal processing hardware and software, significant progress has been made in the field of speech recognition. However, in spite of all of these technological advancements, robots will never be able to equal the performance of their human counterparts in both accuracy and speed, particularly with regard to speaker independent speech recognition. In a nutshell, the process of speech recognition may be broken down into three primary stages: auditory processing, extraction of features, and recognition of classification. The objective of feature extraction is to depict a voice signal by making use of a set number of signal components. This is due to the fact that all of the information contained in the acoustic signal is extremely difficult to process, and part of the information is not pertinent to the identification process. The Northern Goshawk Optimization algorithm, which identifies the distinct characteristics of the speech signal based on the extracted features, is utilised in the process of selecting the features to be used in the analysis. In the end, the speech will be transformed into writing. When compared to the previous work, the new system performs significantly better.