Feedback Cancellation in Public Address systems using notch
Feedback Cancellation in Public Address systems using notch filters Shailesh Kulkarni Vaibhav Mathur Digital Signal Processing-Advanced Topics
Presentation Outline Feedback in audio systems Design of feedback controller – Feedback detection method – Feedback cancellation Implementation and Results – Design of Notch Filter – Graphs and simulation Concluding Remarks 2
Feedback in audio systems 3
Feedback in audio systems • Acoustic coupling between microphone and loudspeaker • Amplified sound at speaker output is propogated back to the microphone and reamplified • Causes oscillation of audible frequencies and high pitch howling sometimes • Can cause physical damage to the audio equipment 4
Feedback Controller IN Feedback Detection notch filter Filterbank OUT 5
Feedback Controller Feedback Detection Works on 1024 point buffer at a time Analyze frequencies with large magnitude Compare them with the harmonics to distinguish between signal and the feedback Candidate frequencies is passed to filter for removal Feedback Cancellation Design of notch filter Changes with every buffer 6
Feedback Detection block 1024 point BUFFER IN 1024 point FFT normalize FFT Find 3 largest Mag. freq. Compare fmax to harmonics Candidate in 3 out of 5 buffers? Fmax = feedback frequency Compare f 2 max to harmonics Candidate in 3 out of 5 buffers? F 2 max = feedback frequency Compare f 3 max to harmonics Candidate in 3 out of 5 buffers? F 3 max = feedback frequency 7
Design of Notch Filter Set NEW notch filter FFB = NEW ? Depth = 3 d. B Width = 1/10 octave Increase depth of existing notch filter Depth = depth + 3 d. B 8
FB detection using Energy Threshold • Alternative way to detect feedback • Deals with relation between energy of the peak and the whole spectrum • Value of Energy threshold determines the sensitivity of the detection 9
Notch Filter We use a parametric IIR filter rz= zero radius rp=pole radius(r) 10
Notch Filter • • Effect of rp – It affects the bandwidth and the depth of the Notch – As rp increases ( < rz ) the notch bandwidth becomes narrower – Also the notch depth decreases[1] We fixed rz to 1 so as to get large notch depth 11
Notch Filter - Simulation 12
Simulations 13
Simulations 14
Performance Metrics RMS as a measure • • – With feedback = 0. 1479 – After filtering = 0. 0114 – Improvement factor = 13 Audio 15
Concluding Remarks • Requirement of better algorithm to detect feedback in the signal for performance improvement • As number of samples per frame increases, we get better resolution in frequency • But we trade delay and complexity 16
References [1] A. F. Rocha and A. J. S. Ferreira, "An accurate method of detection and cancellation of multiple acoustic feedbacks, " Preprints 118 th AES Conv. , Barcelona, Spain, May 2005, Preprint no. 6335 [2] T. van Waterschoot and M. Moonen, "A pole-zero placement technique for designing second-order IIR parametric equalizer filters, " IEEE Trans. Audio, Speech, Lang. Process. , vol. 15, no. 8, Nov. 2007, pp. 2561 -2565 [3] P. A. Regalia and S. K. Mitra, "Tunable digital frequency response equalization filters, " IEEE Trans. Acoust. , Speech, Signal Process. , vol. ASSP-35, no. 1, Jan. 1987, pp. 118 -120 [4] U. S. Patent Number 5245665; www. sabine. com 17
Thank you Questions ? 18
- Slides: 18