Spoken Syllable Recognition Based on Animal Management and Endpoint Detection Algorithm

The task of speech endpoint detection is to distinguish speech segments and non-speech segments from noisy speech signals. They are widely used in speech communication fields such as speech enhancement, speech encoding, and speech recognition. The purpose of this paper is to study the recognition of spoken English syllabic stress in animals by combining animal management and endpoint detection algorithms in order to improve the performance of the speech endpoint detection system in real-time spoken English communication in animal management. This paper uses the frequency domain of the frame signal and the frequency domain of the subband signal to calculate the identification information of the frame signal and the noise frame based on the subband frequency domain distribution probability. The algorithm can update the frequency domain of the noise in all frames including the speech frame by using the identification information, thereby tracking the changes in the syllable frequency domain more accurately, and achieving better recognition of spoken accent syllables. The experiments show that the endpoint detection algorithm has a better overall recognition accuracy for phoneme spectrograms. The actual accuracy has a positive correlation with the segmentation preprocessing effect of the spectrogram. Different phonemes can achieve different recognition effects due to their different pronunciation characteristics. The overall recognition accuracy of all phonemes is about 83%, and the accuracy of spoken accent recognition is about 92%. Key words: Speech Recognition, Animal Management, Endpoint Detection, Spoken English, Accent Syllable Recognition