EURASIP Journal on Audio, Speech, and Music Processing

Table 5 Naturalness MUSHRA average scores and 95% confidence interval

From: A unit selection text-to-speech-and-singing synthesis framework from neutral speech: proof of concept

	Configuration	T=100 bpm	T=50 bpm
S7	V	74 ± 6	70 ± 6
	S7pdLC	41 ± 5	44 ±6
	S7pdC	39 ± 6	43 ± 6
	S7p	36 ± 5	40 ± 6
	MLC	42 ±5	44 ±5
S4	V	69 ± 7	67 ± 7
	S4pdLC	42 ±6	38 ± 6
	S4pdC	39 ± 6	41 ±6
	S4p	35 ± 6	38 ± 6
	MLC	38 ± 6	37 ± 6
S0	V	66 ± 7	70 ± 6
	S0pdLC	44 ±5	38 ± 4
	S0pdC	44 ±5	42 ±6
	S0p	41 ± 6	35 ± 6
	MLC	44 ±6	39 ± 5

Best values achieved by the proposed system in each scenario are in italics

Back to article page