Let Google Voice Search speech better and faster understanding




Google uses a new acoustic model for speech recognition of its Google App on Android and iOS. This would have to be pronounced searches accurately and quickly recognized, even when there is ambient noise.

Google used voice recognition now that the long short-term memory type of recurrent neural networks. This type of network can also be temporal inputs properly classify, process, and to predict when there are long-term dependencies play a role. In the words of Google called the network information longer able to ‘remember’ by using memory cells in the networks and advanced gating mechanisms.

The search engine company cites the example pronounced the word ‘museum’ in English. That word is phonetically spelled / mjuzi @ m /. If the users / u / pronounce the sound production of the / j / and / m / there to be preceded by the movements in the mouth and throat. The RNN could detect this kind of smooth transitions.

Google had to train for these ‘fluid detection’ models for the phonemes to recognize or smallest units of sound, without them separately for each time interval had to make a prediction. This workout models create a series of peaks that represent the successive phonetic units in the speech signal. This enables the model in a position further away from the phonemes in advance and thereby to predict accurately. The model also caused a delay of 300 milliseconds, Google writes. Through further training, the company has to undo know.

Not only is the recognition more accurate and faster, the influence of ambient noise is reduced and requires less computing model. Google published in July, all of the research results, improvements in the recognition of speech.


In: Technology & Gadgets Asked By: [15519 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »