Vocal Music Teaching Pharyngeal Training Method Based on Audio Extraction by Big Data Analysis

Table 2

Statistics of original feature set.

Feature type	Characteristic term	Dimension	Sequence number of item in feature set
Time domain feature	Short-term energy	1	2
	Short-time zero crossing rate	1	1
	Short-term average amplitude difference	1	2
Frequency domain feature	Frequency domain energy	1	5
	Subband energy ratio	4	7-9
	Frame-based spectral centroid	1	11
Cepstrum domain feature	Mel cepstrum coefficient	12	28-40
	Linear prediction cepstrum coefficient	12	15-25
Other features	Loudness range	1	53
	Reflectance	1	58-70