How Speech to text Software works

Things like Siri or google now on our phones rely on speech to text software to work and without them there would be no way to interact with these services. Speech to text software has many other uses besides in our phones and it is being used more and more. What this software does is take recorded speech and determines what was said in that recording. What it needs to function is a microphone to read audio from, and an internet connection. What happens when we ask our phones to do something is the phone’s microphone sends audio data to a server where it is broken down into tiny bits of speech called Phonemes. Phonemes are short distinctive sounds in a language that are all unique. These sounds are usually made up of vowels or combinations of vowels like the “oa” sound in road. There are 44 different phonemes in the English language and a speech recognition program can read these phonemes and use the order and context of them to determine what you are saying. The software uses a similar method to if/else statements we have been doing in class. The program takes a phoneme from the audio of your voice and compares it to the 44 phonemes in the English language. If it finds a match, then it gets the meaning of that bit of speech from what the audio bit was matched to. The program does this for every bit of audio that was recorded until it can determine the words you said. this type of speech recognition is used for many things like automated telephone systems, dictation to a computer, and (as mentioned earlier) for virtual assistants in our smartphones. 

Sources: 

http://scienceline.org/2014/08/ever-wondered-how-does-speech-to-text-software-work/

http://www.lancsngfl.ac.uk/curriculum/literacy/lit_site/lit_sites/phonemes_001/

http://www.technologyguide.com/feature/ios-users-must-try-google-now/google-now-voice-search-screenshot/

http://ios.wonderhowto.com/how-to/siri-exploited-again-bypass-lock-screen-ios-8-protect-yourself-0157749/