Adaptive noise cancellation of coupled transmitter burst noise in speech signals for wireless handsets

This paper describes an application of adaptive noise cancellation (ANC) in the form of an adaptive multiple notch filter [1–3], to remove transmitter burst envelope noise that may be induced in high gain audio circuits of a wireless handset device. We exploit the periodicity of TDMA type systems where the transmitter burst envelope is stationary. Since samples of the transmitter burst envelope are correlated to the additive noise that corrupts the sampled speech, we use this as a reference input to the ANC. We may also obtain the reference input by delaying the primary input to de-correlate the speech. The system was simulated using GSM [4] transmit burst model to demonstrate the noise power reduction as a function of adaptation. Preliminary tests using the GSM transmitter burst model indicate at least −40 dB of noise reduction after ½ second of convergence. The spectral noise characteristics after system convergence are analyzed. We also use a test speech signal to demonstrate the system.