This paper presents the experimental evaluation for recognition of consonant-vowel (CV) units under noise. Noise is one of the common degradation in real environments which strongly effects the performance of speech recognition system. In this work, initially effect of noise on recognition of CV units is studied by using two-stage CV recognition system proposed in our earlier studies. Later spectral processing based speech enhancement methods such as spectral subtraction and minimum mean square error (MMSE) are used for preprocessing to improve the CV recognition performance under noise. Performance of the CV recognition is studied on Telugu broadcast database for white and vehicle noise. Experimental results show that the speech enhancement techniques gives the improvement in the CV recognition performance under noise case.