公开(公告)号：KR1020140099565A

公开(公告)日：2014-08-13

申请号：KR1020130002897

申请日：2013-01-10

Applicant: 미디어젠(주)

Inventor： 이성재 ,  강선미 ,  송민규

IPC: G10L15/01 ,  G10L15/20

CPC classification number: G10L21/007 ,  G10L15/01 ,  G10L15/20

Abstract: Disclosed is an acoustic driving simulator for a virtual performance assessment of an automotive speech recognizer capable of performing an assessment of the automotive speech recognizer without driving a vehicle by a person. For this, the present invention provides an acoustic driving simulator which stores driving noise by driving condition and standard voice of a speaker, converts the standard voice according to the driving condition to apply the Lombard effect to the standard voice, and outputs the driving noise corresponding to a driving condition set from the outside and the voice with the Lombard effect applied. According to the present invention, by improving a conventional method which needs to determine recognition or misrecognition by boarding a vehicle in person or by making a voice recognition command to expensive equipment, speech recognition assessment result similar to real driving can be obtained by simulating the voice command and the generated driving noise indoors, thereby expecting to reduce time and cost burdens.

Abstract translation: 公开了一种用于汽车语音识别器的虚拟性能评估的声学驾驶模拟器，其能够执行汽车语音识别器的评估，而不用人驾驶车辆。 为此，本发明提供了一种声音驾驶模拟器，其通过驾驶状况和扬声器的标准声音来存储驾驶噪音，根据行驶条件转换标准声音，以将伦巴第效应应用于标准声音，并输出相应的驾驶噪音 从外部设置的驾驶状况和应用Lombard效果的声音。 根据本发明，通过改进需要通过亲自登车或通过对昂贵的设备进行语音识别命令来确定识别或误识别的常规方法，可以通过模拟语音来获得与实际驾驶相似的语音识别评估结果 命令和室内产生的驾驶噪音，从而期望减少时间和成本负担。