CAJ | 학술논문

本文在传统发音唇动分析模型的基础上，构建一个发音唇动时空模型．提出了唇动时域特征、空域特性与语音的相关度度量方法，以及融合时空度量的语音唇动一致性检测方法．利用唇宽、唇高与音频幅度变化之间的联系获得语音唇动的时域一致性评分；通过协惯量分析法获得语音与唇部空域特征的初始相关度，并提出了针对语音、唇动自然延时的相关度修订方法；最后将时空上的得分进行融合以判断语音唇动是否一致．初步实验结果表明，对于四种不一致音视频数据，与常用的协惯量方法相比，EER（Equal Error Rate）平均下降了约8．2％．
본문재전통발음진동분석모형적기출상，구건일개발음진동시공모형．제출료진동시역특정、공역특성여어음적상관도도량방법，이급융합시공도량적어음진동일치성검측방법．이용진관、진고여음빈폭도변화지간적련계획득어음진동적시역일치성평분；통과협관량분석법획득어음여진부공역특정적초시상관도，병제출료침대어음、진동자연연시적상관도수정방법；최후장시공상적득분진행융합이판단어음진동시부일치．초보실험결과표명，대우사충불일치음시빈수거，여상용적협관량방법상비，EER（Equal Error Rate）평균하강료약8．2％．
This paper constructs a spatiotemporal lip motion model based on traditional simple pronunciation and lip motion spatial model ,and proposes methods for measuring the correlation degree between voice and the spatial ,temporal characteristic of lip motion .In addition ,a fusion scheme for the spatial and temporal correlation degree is proposed to measure the consistency of voice and lip motion .The temporal consistent score is defined as the correlation between lip shape (height and width) and the speech am-plitude .The Coinertia is used as the initial correlation degree of speech and lip spatial characteristic .Both the spatial and temporal correlation degrees are modified by audiovisual initial delay .Experimental results show that the proposed method reduces EER by about 8 .2% compared to the CoIA method .