Reduction of false hits in speech recognition systems

Embodiments of the present invention improved methods for performing speech recognition. In one embodiment, the present invention includes a method, comprising: receiving a spoken utterance, processing the spoken utterance in a speech recognizer for generating a recognition result determining consistencies of one or more parameters of the component sounds of the spoken utterance, wherein the parameters selected from a group are made of length, intensity and tone, and wherein each component According to the spoken utterance having a respective value of the parameter, and validate the recognition result based on the consistency of at least one of the parameters.