Speaker Recognition/Identification/Verification

说话人识别/认同/验证

In phonetics and computational linguistics, speaker recognition is the acoustic analysis of a speech sample to infer the identity of the speaker. In speaker verification, a sample of speech is acoustically analysed to check a claimed identity against a stored reference sample stored in the computer. In speaker identification, a speech sample from a known speaker is compared to one obtained from an unknown speaker, to determine whether the same person is involved. This technique has been most commonly used in criminal cases, where the analysis of spectrograms (as in voiceprinting) has been used to investigate whether the speaker in a tape-recording is the same as a suspect.

语音学和计算语言学对某一言语样品作声学分析，以此推断说话人的身份。说话人验证是对某一言语样品作声学分析，将一设定的身份与计算机内存储的参照样品相比较和验证。说话人认同是将一已知说话人的言语样品与一未知说话人处获得的样品相比较，以决定两者是否为同一个人。这项技术最常用于犯罪侦破，频谱的视觉分析（声纹）可用来调查录音上的说话人是否就是犯罪嫌疑人。