Speaker adaptation based on regularized speaker-dependent eigenphone matrix estimation

EURASIP Journal on Audio, Speech, and Music Processing

Table 4 Average tonal syllable recognition rate (%) after eigenphone-based speaker adaptation using group lasso

λ ₂	Number of adaptation sentences
	1	2	4	6	8	10
60	52.56	53.36	56.84	58.06	59.78	60.85
	(0.07)	(0.0)	(0.0)	(0.0)	(0.0)	(0.0)
90	53.84	54.51	56.90	58.37	59.86	60.45
	(0.34)	(0.02)	(0.0)	(0.0)	(0.0)	(0.0)
120	54.22	55.77	57.03	58.06	59.63	60.34
	(0.65)	(0.12)	(0.01)	(0.0)	(0.0)	(0.0)
150	54.26	55.33	56.99	57.97	59.30	60.30
	(0.84)	(0.32)	(0.03)	(0.0)	(0.0)	(0.0)

The number of eigenphones (N) was fixed to 100. λ₁=λ₂=0, and λ₃ was varied between 10 and 150. The average column sparsity of the eigenphone matrix is shown in parentheses.