What Is Continuous Speech Recognition?

Continuous voice recognition refers to the recognition of continuous audio streams (that is, voices directly entered by speakers, or audio signals from telephone or other audio and video fields), and automatically converts audio information into text.

1.Hidden Markov Model

Hidden Markov model is a statistical model that is used to describe a Markov process with hidden unknown parameters. It is widely used in speech recognition. 1 The system first generates a speech model from a large amount of text, then extracts the acoustic features, and obtains the recognition result after Viterbi decoding.

2.Methods based on convolutional neural networks

The convolutional neural network analyzes local features through a convolutional device, strengthens the extracted feature robustness through the aggregation layer, and finally builds a model through the full network layer to get the final classification result. The convolutional neural network observes local features through the convolutional layer, and finally obtains the output probability through the information integration of the entire network layer, which has better physical meaning than the deep neural network.

1. In the field of security, relevant departments have put forward corresponding requirements in combination with related businesses; in the field of education, the Mandarin test and spoken assessment of large groups of people urgently need objective and automatic assessment techniques;

2. In the field of telecommunications, domestic and foreign speech recognition technologies and departments have entered the Chinese market;

3. In the embedded markets such as mobile phones and car navigation, the demand for voice recognition technology is also increasing;

4. In the field of human-computer interaction, voice companion and voice search of mobile terminals are widely used.

Therefore, speech recognition technology, as a very important human-computer interaction technology, has a very broad prospect. ^[1]