In this paper, a novel method to reduce additive time-varying noise is proposed. Unlike the previous methods, the proposed method requires neither the assumption about noise nor the estimate of the noise statistics from any pause regions. The enhancement is performed on a band-by-band basis for each time frame. Based on both the decision on whether a particular band in a frame is speech or noise dominant and the masking property of the human auditory system, an appropriate amount of noise is reduced in time-frequency domain using modified spectral subtraction. The proposed method was tested on various noisy conditions: car noise, F16 noise, white Gaussian noise, pink noise, tank noise and babble noise. On the basis of segmental SNR, inspection of spectrograms and MOS tests, the proposed method was found to be more effective than spectral subtraction with and without pause detection in reducing noise while minimizing distortion to speech.