Automatic grading method and automatic grading equipment for read questions in test of spoken English

The present invention provides a method and apparatus for reading English spoken automatic scoring. The method comprises: preprocessing of the input speech, the pretreatment process comprising framing; extracting speech features from the voice pretreated; linear network grammar and acoustic models built read text, speech feature vector sequences are forced alignment to give each phoneme dividing point information; divided for each of the phonemes point information, calculating posterior probability of each phoneme; phoneme posterior probability based on a multidimensional score feature extraction; the manual scoring information and feature score, training the support vector regression non-linear regression model to take advantage of the non-linear regression model for spoken English reading score. Experts scoring data using a scoring model training, to ensure that the statistical results from the scoring machine will not deviate manual scoring, so as to achieve a high degree computer simulation of the expert score.