Highlight of words in audio signal is a part of large project. This task is the first step of speech recognition systems also known as speech-to-text systems.
We needed to create algorithm which can be able to highlight words in audio stream. Algorithm should split the signal on parts, each of them represents one of words in English.
Audio signal has many characteristics in time and frequency area. In this task was reviewed energy and entropy as key characteristics to split signal. The serie of experiments gave ability to determine entropy as the best characteristic to split audio signal on words.
Click on image to view it on full size