Currtnly, there are services like Siri, Alexa, Google Assistant, LINE Clova, and many other audio input devices. Listen to time, weather, shopping, just to be able to do with just "voice", it is a science fiction world.
Even so, how is this speech recognition technology realized by what mechanism? The history was introduced in IBM 's Web media infinity.
According to the article, the company started researching speech recognition from the 1960's. One of the typical achievements is telephone conversation speech recognition, which seems to be a system for recognizing natural conversation between humans for the speech of a low-quality telephone line.
It seems that it is made up of two parts, acoustic model modeling which sounds the input speech signal corresponds to, and language model modeling natural word sequence. Deep learning has been introduced to this in recent years, its accuracy has dramatically increased. It seems that it is possible to recognize the same degree of speech as human beings.
According to the forecasts of relevant organizations, speech-based shopping is expected to grow from the current 2 billion dollars to more than 40 billion dollars in 2022. Perhaps it may not be using a keyboard or smartphone at that time.
If you are interested in the history and current state of speech recognition technology, please enjoy continuing reading.
