Sophisticated Adaptive Signal Processing Systems for Howling Canceller

本项目研究有效消除啸叫的数字信号处理系统。消除啸叫的一个基本思想是使用自适应陷波滤波器来跟踪和抑制啸叫信号，因为它近似于频率未知的正弦波。本项目的主要成果如下：(1)实际的啸叫信号由于相位抖动而不具有线谱，而具有带通谱。为了有效地抑制这种啸叫信号，需要检测其宽度并确定自适应陷波滤波器的带宽。然后提出了一种利用二阶自适应陷波滤波器检测带通信号的通带边缘频率的方法。该方法于2006年3月在IEICE电路与系统技术小组研讨会上发表。现在它正在准备提交IEICE交易。(2)通过整合前三年的研究，使去年的啸叫信号实时消除成为可能，这是本项目的最终目标。嚎叫消除器以软件系统的形式在个人计算机上实现，用声音芯片作为输入/输出设备。由于自适应系数的收敛性还不完善，存在很小的残余啸叫。然而，当不应用这种消除器时，由于大啸叫，麦克风放大器不能很好地工作。因此，所设计的啸叫消除器是有效的。现在更多的研究是在专利申请方面提高收敛性。

This project investigates digital signal processing systems for canceling howling efficiently. A fundamental idea to cancel howling is to use adaptive notch filters to track and reject howling a signal because it is approximately a sinusoid whose frequency is unknown. Main results of this project are as follows ;(1)An actual howling signal does not have a line spectrum but a bandpass one due to phase jitter. In order to reject such a howling signal efficiently, it is required to detect its width and to determine the bandwidth of an adaptive notch filter. Then a method for the detection of passband edge frequencies of a bandpass signal by using a second-order adaptive notch filter has been proposed. This method was presented IEICE Technical Group Workshop on Circuits and Systems on March 2006. Now it is in preparation to submit IEICE Transactions.(2)By integrating the studies in early three years it became possible to cancel howling signals in real time in last year, which is the end goal of this project. The howling canceller is realized as a software system on a personal computer, whose sound chip is used as an I/O device. Since the convergence property of an adaptive coefficient is not still perfect, very small residual howling exists. However, when this canceller is not applied, microphone amplifiers cannot work well due to large howling. Therefore, the designed howling canceller is effective. Now more studies to enhance the convergence property is doing toward patent filings.