Discrete Word Speech Recognition Using Hybrid Self-adaptive HMM/SVM Classifier This research addresses independent speaker’s discrete word speech recognition (DWSR) using hybrid Self-adaptive Hidden Markov Model/Support Vector Machine (SA-HMM/SVM) classifier. Our proposed method includes two main units: preprocessing unit, and classification unit. The first unit tries to frame the speech wave into proper segments and extract time-frequency relevant features in a way to maximize relative entropy of time-frequency energy distribution among segments, and the second unit classifies words within the proper classes. To fulfill this goal, SA-HMM calculates word’s likelihood to each existing class correspondently, and finally Support Vector Machine (SVM) classifies it by using all classes’ likelihood as an input vector. To validate our proposed method, we test it within our IAUM dataset which contains Persian digits uttered by Persian speakers. Comparing the results with the outcomes of a similar method based on the original HMM shows around 1.2% improvement Discrete Word Speech Recognition, Local Orthogonal Discriminate Bases, Hybrid SVM/Self-adaptive HMM classifier 80 90 Saeid Rahati Quchani سعيد راحتي قوچاني rahati@mshdiau.ac.ir No Kambiz Rahbar كامبيز رهبر No