Table 1 Results for a simulated STNO neural network with N = 400 nodes.

From: Role of non-linear data processing on speech recognition task in the framework of reservoir computing

Filter

Training

Testing

Overfitting

WSR (std)

MSE (std)

WSR (std)

MSE (std)

MSEtest/MSEtrain

Cochleagram

100.0% (0.0%)

0.0133 (1.1 · 10−4)

99.6% (0.8%)

0.0193 (2.3 · 10−3)

1.451

MFCC

99.8% (0.1%)

0.0259 (1.1 · 10−4)

99.2% (1.0%)

0.0283 (8.2 · 10−4)

1.093

\(\Re \)(Spectro)

75.7% (0.6%)

0.0708 (1.1 · 10−4)

65.2% (4.2%)

0.0738 (7.3 · 10−4)

1.042

Spectro HP

99.2% (0.2%)

0.0362 (9.9 · 10−5)

98.6% (1.9%)

0.0384 (6.5 · 10−4)

1.061

Back to article page