当前位置:网站首页>Voiceprint Technology (III): voiceprint recognition technology

Voiceprint Technology (III): voiceprint recognition technology

2022-06-25 09:05:00 u013250861

3.1  Voiceprint recognition : The core of voiceprint technology

3.1.1  Name and concept

A broad sense , Voiceprint technology is a broad concept , It contains many different technologies and applications . Of all these technologies , Voiceprint recognition technology is the basis of other technologies . No matter it's the 5 The voiceprint segmentation and clustering technology will be introduced in chapter , Or no 6 This chapter will introduce the speech synthesis based on voiceprint 、 Voice separation and voice activity detection , Can not be separated from the cooperation with the voiceprint recognition model , The voiceprint recognition model can be pre trained (pre-trained), It can also be joint training (joint training) Got . therefore , This chapter is also the most important part of this book 、 The core chapter .

Voiceprint recognition , Also known as speaker recognition , It corresponds to several kinds of expressions in English , for example voice recognition、speaker recognition、voiceprint recognition、talker recognition And so on are the same concept , That is to say, the pronunciation of different speakers , The technique of distinguishing according to the identity of the speaker .

Here we should pay attention to the combination of voiceprint recognition and speech recognition (speech recognition) Technology . Speech recognition is a technology that recognizes speech signal as text content , In most cases, I don't care who the speaker is , And need to be robust to different speakers' voices [1]. And voiceprint recognition technology —— Especially text independent voiceprint recognition technology —— On the contrary , It is necessary to identify the speaker's identity in different text content . From this we can see that , In a sense, speech recognition and voiceprint recognition are regarded as two mutual “ orthogonal ” The problem of : Speech recognition hopes to filter out the information related to the identity of the speaker from the signal , Keep only text information ; Voiceprint recognition hopes to filter out the information related to the text from the signal , Only the identity information of the speaker is retained .

1.1 As mentioned in section , Everyone's hair

原网站

版权声明
本文为[u013250861]所创,转载请带上原文链接,感谢
https://yzsam.com/2022/176/202206250736417820.html