Speech Recognition Using Weighted Projection-Based Likelihood Measure of Cepstral-Time Features and Energy-Time Feature