Speech Detection and Comparison using Different Feature Extraction Method for Correct Verification
Shrikant Upadhyay, Binod Kumar, Rajan Singh, Md. Irfan Alam, Barkha Kumari, Binita Roshima Hinz · 2024
Mental status through physical appearance identification is not possible as it does not give complete status of any human. Using voice and considering quality extraction of talked it is quite easy to predict the mental behavior and internal variation that will help the doctor related with medical treatment that support them to give proper treatment to person. Hindi is the natural language of India and large community of population talk and speaks this language to exchange and communicate. Various inputs like specific feature, images and vital report of serious person and it is one of the important facts sources for deep training or gaining systems. In this research paper we try to identify the speech signal under different scenario like gun noise, car noise etc. following different feature extraction approaches and compare it for its proper verification. This verification may help to enhance the quality of speech in different adverse scenario and make the speech detection easy. Here, the comparison has been done with various feature extraction method and the results reflects that MFCC has good efficiency of 95% and with a low error rate of 2% compared to other approach. For analysis Hindi voice sample were used for the experimental analysis for 50 samples.