Whisper speech feature extraction method and system

Abstract

The invention discloses a whisper feature extraction method which is characterized by comprising the following steps of (1) whisper speech auditory spectrum feature representation; (2) feature dimension reduction and robust performance analysis; the feature dimension reduction and robust performance analysis include three contents: (a) the low-dimensional topological structure features are extracted from the high-dimensional auditory spectrum; (b) topological structure feature time sequence analysis is performed; and (c) topological structure feature stability analysis is performed; and (3) acoustic model optimization is performed; the training mechanism of performing passive learning and then active learning is adopted in acoustic model optimization so as to enhance the adaptive performance of the model. According to the whisper speech feature extraction method, dimension reduction is performed on the auditory perception spectrum features to obtain the topological structure features, and the distinguishability of the whisper speech features is strengthened by strengthening the time sequence weight of the features. The distance between the feature vectors of different meanings is maximized through two target functions, and the distance between the feature vectors of the same meaning is also minimized so as to enhance the robustness of the features.

Description

technical field [0001] The invention relates to speech signal processing technology, in particular to an ear speech feature extraction method and system. Background technique [0002] The process in which the human ear receives speech and understands semantics through auditory mechanism and nervous system processing is called speech perception, which is the process of "hearing" and "understanding". Among them, "understanding" is the key. The objects of "understanding" include semantics, personality information, social information, etc., and understanding semantics is the most important and critical issue in speech perception. It is called speech recognition in speech signal processing. Although speech recognition has achieved relatively satisfactory results under laboratory conditions, it is still difficult to perform speech recognition under conditions such as special pronunciation methods, strong noise backgrounds, and voice personality diversification. At present, speech...

AI Technical Summary

Problems solved by technology

The main reason is that ear speech has some important characteristics: ear speech is breathy and low in energy, and it is usually 15-20dB weaker than normal speech; the excitation source of ear speech is noise, and the vocal cords do not vibrate, so ear speech has no fundamental frequency; The channel transfer function changes, adding extra poles and zeros, resulting in a flatter spectrum of ear speech than normal speech, greater attenuation in the low frequency band, and a shift in the spectrum; the average duration of ear speech increases compared with normal speech, and the voice onset time of voiced consonants is different from that of normal speech. There are significant differences in pronunciation; whispering pronunciation is more likely to be affected by the speaker's psychology and environment, the Lombard effect is more obvious, the voice characteristics change greatly, and the randomness is strong

Therefore, ear speech recognition, especially the representation of ear speech features used for recognition, cannot copy the way of normal speech

Images