Decision tree based speech recognition

摘要: A method ( 200 ) is described for creating decision trees processing a sampled signal indicative of speech. The includes providing model sub vectors 220 from partitioned statistical speech models phones, the comprising mean values and associated variance values. then provides statistically analyzing 230 to provide projection indicating directions relative maximum between thereafter calculating 240 vectors. selecting potential threshold 260 step applied, being determined analysis range Finally 270 effected tree having decisions divide into groups, groups leaves tree. are based upon selected values, by change in said There also recognition 300 that uses created ).