Noise Spectrum Estimation in Adverse EnvironmentsImproved Minima Controlled Recursive Averaging 恶劣环境下的噪声谱估计：改进的最小控制递推平均法.pdfVIP

466 IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL. 11, NO. 5, SEPTEMBER 2003 Noise Spectrum Estimation in Adverse Environments: Improved Minima Controlled Recursive Averaging Israel Cohen Abstract—Noise spectrum estimation is a fundamental compo- A useful noise estimation approach, known as the minimum nent of speech enhancement and speech recognition systems. In statistics (MS) [12], is to track the minima values of a smoothed this paper, we present an improved minima controlled recursive power estimate of the noisy signal, and multiply the result by a averaging (IMCRA) approach, for noise estimation in adverse environments involving nonstationary noise, weak speech com- factor that compensates the bias. However, the variance of this ponents, and low input signal-to-noise ratio (SNR). The noise noise estimate is about twice as large as the variance of a con- estimate is obtained by averaging past spectral power values, using ventional noise estimator [12]. Moreover, this method may oc- a time-varying frequency-dependent smoothing parameter that is casionally attenuate low energy phonemes, particularly if the adjusted by the signal presence probability. The speech presence minimum search window is too short [4]. These limitations can probability is controlled by the minima values of a smoothed periodogram. The proposed procedure comprises two iterations be overcome, at the price of significantly higher complexity, by of smoothing and minimum tracking. The first iteration provides adapting the smoothing parameter and the bias compensation a rough voice activity detection in each frequency band. Then, factor in time and frequency