Skip to main content

Table 2 Best features for each of the five thresholds.

From: A Decision-Tree-Based Algorithm for Speech/Music Classification and Segmentation

Threshold type

Features

Extreme speech

(1) 9th MFCC (mean val. of diff. mag.)

 

(2) Energy (std)

 

(3) 9th MFCC (std of diff. mag.)

 

(4) LSTER

Extreme music

(1) High Band Energy Ratio (mean value)

 

(2) Spectral rolloff point (mean value)

 

(3) Spectral centroid (mean value)

 

(4) LSTER

High probability speech

(1) Energy (std)

 

(2) 9th MFCC (mean val. of diff. mag.)

 

(3) Energy (mean val. of diff. mag.)

 

(4) Autocorrelation (std)

 

(5) LSTER

High probability music

(1) Energy (mean val. of diff. mag.)

 

(2) Energy (std)

 

(3) 9th MFCC (std of diff. mag.)

 

(4) Autocorrelation (std of diff. mag.)

 

(5) ZCR (skewness)

 

(6) ZCR (skewness of diff. mag.)

 

(7) LSTER

Separation

(1) Energy (std)

 

(2) Energy (mean val. of diff. mag.)

 

(3) Autocorrelation (std)

 

(4) 9th MFCC (std of diff. mag.)

 

(5) Energy (std of diff. mag.)

 

(6) 9th MFCC (mean val. of diff. mag.)

 

(7) 7th MFCC (mean val. of diff. mag.)

 

(8) 4th MFCC (std)

 

(9) 7th MFCC (std of diff. mag.)

 

(10) Autocorrelation (std of diff. mag.)

 

(11) LSTER