In this work, we present a low-complexity single-ended objective intelligibility measure for noisy speech based on statistics computed from auditory modulation features. The proposed measure is obtained in two steps. First, we compute several statistics of auditory representation of corrupted speech. Next, a support vector regressor (SVR) is used to map these statistics to an overall intelligibility score. The SVR is trained using subjective intelligibility data. The proposed measure shows high performance in predicting intelligibility of speech corrupted with additive noise.