In this paper, a novel method for beginning of utterance detection is proposed for low complexity ASR systems. Assuming MFCC calculations in the ASR front-end, the additional computational load due to the algorithm is negligible. The algorithm makes use of the delay between the MFCC calculation and decoding process, which is typical in front-ends with feature normalization. The main steps of the algorithm involve LDA projection of MFCC features, mean calculation over the projected features, simple implicit SNR estimation and weighting of the decision statistics according to the estimate. Our experimental results show that high performance is obtained down to fairly low SNR conditions as the beginning of utterance detection starts to fail in a safe way at about 5 dB SNR. These properties make the algorithm an attractive choice for low complexity ASR engines.