A novel approach towards voice-based video content search

This paper describes challenges and solutions for building a successful voice-based video content search system. People have been typing to search queries online for the past several decades. But it is easier and more convenient to use voice search than typing the whole text. Voice based search is dominating the market since last few years. With the expanding dominance of smartphones this voice search also steadily growing. We have implemented voice search instead of web search, specifically it is voice-based video content search. A particular word has been provided as voice input in a video file and it is tried move directly to that point of the video. The ffmpeg tool has been used to convert the video file into corresponding file format. Speech recognition module has been used to convert audio file into text. The required attributes from speech recognition module provide us all the possible words similar to the given audio input, so that we won't miss any possible word or sentence. The texts are divided and saved in different locations in an array for every interval of 3 seconds. To access the voice input word, the index value has been multiplied to the duration to get the timestamp where the input word lies. Then the user can select the timestamp shown on the screen and move to the intended location of the video.