In view of the problem that the combination of acoustic features (prosodic feature and MFCC feature) is not ideal for the classification and recognition of emotional speech, a cascade classification method for emotional speech recognition that combines acoustic features with emotional speech PAD data is proposed.First, the acoustic features of emotional speech are extracted, and the features are subject to separate recognition and combined identification, and the optimal characteristic set is established by comparison.Then the acoustic feature combination and emotional speech PAD data are combined to determine the emotion type of the input speech step by step.The result of this method is better in TYUT2.0 emotional speech database.The recognition rate of sentiment classification is 15.4% higher than that of traditional acoustic features.