改进MFCC算法在朱鹮鸣声特征提取的应用

Applications of improved MFCC algorithm in crested ibis chirp recognition

基于人耳听觉模型的MFCC参数具有很好的抗噪性,但是其在中高频区域计算精度不足。针对朱鹮鸣声频带较宽、总体频率范围较广的特点,在提取MFCC特征参数时先对幅度谱进行平滑,然后采用IMFCC,Mid MFCC,MFCC相结合的算法,对朱鹮鸣声进行特征提取,以提高在中高频区域中的识别精度。实验结果表明取得了较好的效果。

MFCC parameters are based on the human auditory model and have great noise immunity, but its lack of accuracy in the calculation of the high frequency region. Crested ibis song has wide band and a wide frequency range, so in the first of MFCC parameter extraction amplitude spectrum smoothing, and then using IMFCC, MidMFCC, MFCC combining algorithm, the crested ibis song for feature extraction to improve recognition accuracy in the high frequency region. The experimental results show that it achieved good results.