A GTCC-Based Underwater HMM Target Classifier with Fading Channel Compensation

Table 3

Classifiers performance comparison.

Performance conditions	Prototype HMM classifier (%)	Success rates (%)
		Artificial neural network classifier	Euclidean distance classifier
Without ambient noise	93	89	86
Signals obscured by self-noise
Without spectral subtraction and HMM trained without self-noise signals	47	41	38
Without spectral subtraction and HMM trained with self-noise-affected signal	74	67	63
With spectral subtraction and HMM trained with non-self-noise signals	85	79	68
Signals obscured by Rayleigh fading
Without Rayleigh fading compensation and HMM trained with nonfaded signals	49	44	38
Without Rayleigh fading compensation and HMM trained with faded signals	83	78	75
With Rayleigh fading compensation applied
Autoregressive model with linear interpolation	89	84	81