CAJ | 학술논문

为了提升VOIP的通话质量，提高回声抵消系统中双端通话检测的准确性，本文提出一种新的双门限双端通话检测算法。新算法首先利用小波阈值去噪算法对接收的各信号进行去噪处理；然后对去噪处理后的小波子带信号进行动态能量阈值比较，识别出静音帧、活跃帧和发声帧；再对活跃帧与静音帧进行基音周期比较，以判断活跃帧是处在近端发声阶段，还是回声路径的变化。仿真实验证明，新的双门限检测算法表现良好。
위료제승VOIP적통화질량，제고회성저소계통중쌍단통화검측적준학성，본문제출일충신적쌍문한쌍단통화검측산법。신산법수선이용소파역치거조산법대접수적각신호진행거조처리；연후대거조처리후적소파자대신호진행동태능량역치비교，식별출정음정、활약정화발성정；재대활약정여정음정진행기음주기비교，이판단활약정시처재근단발성계단，환시회성로경적변화。방진실험증명，신적쌍문한검측산법표현량호。
This paper mainly presents a new algorithm for double-talk detection to improve the quality of VOIP calls and its accuracy in echo cancellation system. This new algorithm firstly de-noises the voice signal via wavelet threshold de-noising algorithm, identifies silent frames, active frames and voice frames by comparing dy-namic energy threshold for wavelet sub-band signals after de-nosing. Then the active frames could be determined in a double-talk stage, or changes in the echo path via comparing the pitch determination of active frames and silent frames. Simulation results show that the new algorithm performs well.