Skip to main content

Table 5 The impact of changing the lower and higher cutoff frequencies of the log-mel spectrogram on the performance of each detector for normal running speech signals

From: A CNN-based approach to identification of degradations in speech signals

Detectors flow[Hz] fhigh[kHz]
  0 300 700 2500 4.3 11 15 Nyquist
Noise 0.95 ±0.01 0.92 ±0.01 0.94 ±0.00 0.91 ±0.01 0.99 ±0.00 1.00 ±0.00 0.99 ±0.00 0.95 ±0.01
Distortion 1.00 ±0.00 1.00 ±0.00 0.99 ±0.00 0.99 ±0.00 0.99 ±0.01 1.00 ±0.00 0.99 ±0.00 1.00 ±0.00
Reverberation 0.99 ±0.00 0.99 ±0.00 0.99 ±0.00 0.99 ±0.00 0.90 ±0.01 0.90 ±0.01 0.97 ±0.01 0.99 ±0.00
  1. The results are in the form mean AUC ±95% confidence interval