艺术嗓音歌声客观评价初探

目的探讨客观评价艺术嗓音歌声的方法。方法对48名声乐专业青年大学生录制专业训练歌声信号,提取歌声平均能量、频率误差、音域误差作为评价参数,使用神经网络方法和多元线性回归方法客观评价歌声质量,并与资深专业教师的主观评价进行比较。结果客观评价歌声质量的方法中,神经网络方法误差在4%之内,而线性回归方法误差在6%之内,前者较优。结论神经网络方法利用评价参数能正确客观评价歌声质量,有助于科学地指导选拔和训练艺术嗓音人才。

Objective Evaluation of the Voices of Artistic Singing Voice

Objective To discuss how to evaluate voices of artistic singing objectively.Methods The singing voices were recorded from 48 young music students of singing.The characteristics of singing,such as average energy,errors for frequencies and vocal nange were assessed.Neural network analysis and linear regression were used to evaluate the singing voices objectively.The results were then compared with the subjective evaluations by experienced professionals.Results The errors yielded by neural network analysis were within 4%,and that by linear regression was within 6%.The former seemed more precise than linear regression to evaluate singing voices.Conclusion Neural network analysis can be used as an objective instrument to evaluate singing voices.This can be helpful to select and train professional singers.