Skip to main content

Table 2 SER, error per class and average class error for the 1BLSTM RNN classifier on the test partition for different frontend configurations (Chr, chroma; Δ,ΔΔ, 1st and 2nd order derivatives)

From: Multiclass audio segmentation based on recurrent neural networks for broadcast domain data

FeatsSERClass error (%)Avg
  muspsmsn 
64 bands18.1818.5432.4332.4835.7629.80
80 bands17.7018.1931.3331.4134.9128.96
96 bands17.9320.6830.8432.0934.2529.46
64 bands + Chr16.9718.8330.8829.9232.7628.10
80 bands + Chr17.8919.7732.2329.5533.9228.87
96 bands + Chr17.6519.7530.6831.6233.6628.93
64 bands + Chr + Δ,ΔΔ16.6117.4629.9329.2632.6027.31
80 bands + Chr + Δ,ΔΔ16.2516.8230.0026.7532.0726.41
96 bands + Chr + Δ,ΔΔ16.4617.3829.9227.9832.7027.00