Skip to main content

Table 5 Calculation results in the noise matching condition when IRM is the training target

From: Research on monaural speech segregation based on feature selection

Feature and feature group

STOI

PESQ

SDR

SNR

AMS

0.7554

2.2546

8.5116

8.4045

AMS + GF

0.8218

2.5484

10.8552

10.5734

AMS + GF + MRCG

0.8289

2.5875

11.1200

10.8836

AMS + MFCC

0.8210

2.5083

10.5937

10.3795

AMS + MFCC + GF

0.8277

2.5638

10.9731

10.7261

AMS + MFCC + GF + MRCG

0.8324

2.6112

11.1327

10.8991

AMS + MFCC + MRCG

0.8303

2.5799

11.0467

10.8180

AMS + MRCG

0.8270

2.5666

10.9469

10.7455

GF

0.8190

2.5563

10.8528

10.3649

GF + MRCG

0.8266

2.5774

10.9900

10.7394

MFCC

0.8178

2.4870

10.3623

10.0485

MFCC + GF

0.8260

2.5636

10.9189

10.5575

MFCC + GF + MRCG

0.8291

2.5972

11.1015

10.8133

MFCC + MRCG

0.8271

2.5767

10.9487

10.6630

MRCG

0.8227

2.5593

10.8100

10.5759

  1. Bold data represents the optimal value of each feature group for this indicator