Research on the Construction of Human-Computer Interaction System Based on a Machine Learning Algorithm

Table 1

Spectral characterization.

Extract characteristics	Advantage	Shortcoming
MFCC	Popular features.	Poor noise resistance.
LPCC	Helps to capture the voice perception of the human ear.	For different emotions (especially anger and sadness), the coefficient values usually overlap.
ZCR	Delta and double-delta values can improve recognition accuracy.	The ZCR value tends to vary greatly, depending on the amount of noise present.
Shimmer	Indicates common features of voice content.	Emotions such as anger and disgust often exhibit similar jitters and flickers.
LFPC	The observed LFPC value is not relevant, so the diagonal covariance of its value can be used as the feature input of the classifier.	Most studies only compare LFPC with MFCC and LPCC. The nonlinear changes of the speech signal are not considered.
DSCC	Simple calculation.	Anger and disgust often exhibit similar jitters and flickers.