Spectrogram Analysis and Text Conversion of Sound Signal for Query Generation to Give Input to Audio Input Device

Abstract

AbstractThe world is being reshaped by Natural Language Processing. Audio inputs are used in modern electronics. Different types of people supply input to the system in their native language. The system accepts the person's speech, processes it, and responds accordingly. Cooking is a huge problem for a variety of people, including the elderly, those who are confined to their beds, and those who have a specific sort of handicap, such as those who are unable to use their hands and require assistance at all times. To help these people reach their full potential, an audio input device for giving cooking instructions to a cooking system has been proposed in this paper. The gadget takes the user's spoken English language as input, converts it to text using deep learning algorithms, and generates instructions with the help of context-aware words extracted from the recorded audios to send the instruction to the cooking device. To analyses, the audio signal for user authentication is a challenging task due to gaps and pauses between spoken characters, and existing noise in the environment. As a result, the audio input device developed for kitchen systems must analyze the audio input signal to create a more secure environment for authenticated users. As a result, the objective of this paper is to analyze the audio input signal captured in real-time and process the accepted signal to convert into the text to generate instructions for a larger system. The sound signals captured in the real environment are analyzed with Mel spectrogram, MFCC spectrogram, and PRAAT software. The sound signal is processed with the help of a natural language toolkit to generate instructions.KeywordsMel-spectrogramAudio signalMFCCTokenization