One of the most relevant current research fields related to Computer Science is voice recognition. Direct communication between man and the computer is intended, without the need to transcribe the information through a keyboard or other intermediate information media.
Voice or word recognition devices usually try to identify phonemes or words within a very limited repertoire or vocabulary. A phoneme is a single sound or unit of spoken language.
A system capable of recognizing, say, 7 words, what it does when it detects a sound is to extract characteristics or physical parameters inherent to said sound, and compare them with the parameters (previously memorized) of the 7 words that it is capable of recognizing. If, as a result of the comparison, it is identified as corresponding to one of the 7 words, the word identifier binary code is transmitted to the device buffer. If the sound is not identified, this circumstance is indicated to the user (by lighting up a light, for example) so that the user emits the sound again.
There are two types of voice recognition units:
In these systems it is necessary to subject the device to a learning or programming period, after which it can recognize certain user words. In the learning period, the system retains or memorizes the characteristics or peculiarities of the sounds emitted by the speaker, and which it will then have to identify.
These systems are more widespread, but the vocabulary they recognize is usually very limited. The parameters of the words that identify them are already memorized when purchasing the unit. They are used, for example, to define the movement of certain types of robots. In this case, the operator verbally gives commands chosen from a very limited repertoire, such as: stop, go, up, down,… When the unit picks up a sound, it checks if it corresponds to one of those in the repertoire. In case of identification, the necessary information is transmitted to the central computer for the execution of the program that starts and controls the required action.