This paper proposes a new phone-based Putonghua pronunciation quality assessment algorithm. This work is part of a project aimed at developing an automatic system that can assess the quality of Chinese Putonghua pronunciation. We bring up a new algorithm based on classical log-posterior probability algorithm by weighting each phone of Putonghua to calculate machine score. Experiments show that the new method improves the correlation between machine and human scores, which went up from 0.666 to 0.707. Keywords: speech recognition; pronunciation quality assessment; log-posterior probability; phone weigh