Font Size: a A A

Voice Feedback Suppressor

Posted on:2007-06-22Degree:MasterType:Thesis
Country:ChinaCandidate:C WangFull Text:PDF
GTID:2208360185972990Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Aim at the problems of reverberation, echo and howling etc of public-address system, the paper research these common solutions :parameters equalizer, feedback canceller and frequency shifter at current public-address processors, and analyze the design of the most advanced products at home and abroad, finally put forward new application-oriented solutions.We find that when the distance between microphone and speaker is far the performance of feedback canceller decreased, we decide the phenomenon is induced to the spreading delay of voice, so we design a new adaptive filter which the used orders can slide as a whole, so the performance is no affected by the distance, we use this method at adaptive equalizer which the filter' s factors won' t be changed at normal work. In the acoustic echo canceller design of hand-free cell and communication meeting, a A sample delay is used to decrease the orders of adaptive filter, its purpose is also to compensate the spreading delay from speaker to microphone, but it demand the distance is constant; the order-sliding filter is flexible to get the delay between speaker and microphone when tracking system, its defect is that the judge of sliding is complicated.To improve the capacity of simulating system at dynamic environment and to think about the tracking capacity of filter is only demanded to have, not demanded to be very quick, so give a chance to apply the multi-step-adapting filter. the multi-step-adapting filter is to change all factors at several sample time not one sample time, experiments prove that it won' t decrease the tracking capacity of filter to frequency spectrum of input signal, and the tracking speed become slower than normal filter and faster than block filter. In fact, the block filter has the fastest speed of calculation and should be feasible to improve the filter' s orders, but in common application the feedback voice and voice are mixed together at most time, so the filter don' t have adequate time to finish the tracking. Multi-step-adapting have a limited value to improve filter' s orders because it only decrease the orders that need to be adapted and not decrease the orders that need to calculate convolution.To delete the noise induced by filter' s tracking, a double-filter structure is used, the two filters' factors will be set to zeros every a short time, so it can track dynamical system well and be used to delete white noise and howl inhibiting. The double-filter structure is different with master-slave or front-back structure. Master-filter' s factors are constants during normal work and slaver track the system, when the performance of slaver is better than that of master the slaver' s factors will be copied to the master, (the front-back structure resemble this). the double-filter structure judge that the performance of filter will become bad after a long time tracking, so its factors should be reset periodly to delete the influence of previous tracking, the two filters must keep tracking at any time to make the outputs plane. The techniques used to delete noise can be divided into one channel and two channels. One channel include Winner filter, Kalman filter and frequency spectrum canceller. Frequency spectrum canceller demand the frequency spectrum of noise is stable for the evaluated value...
Keywords/Search Tags:double-tracking adaptive filter, order-sliding adaptive, acoustic feedback inhibitor, frequency shifter, howling
PDF Full Text Request
Related items