Text-to-speech system for low-resource language using cross-lingual transfer learning and data augmentation

EURASIP Journal on Audio, Speech, and Music Processing

Table 6 Number of semitones of pitch shift (PF), or ratio of speed of the new speech to speed of the original speech (SF), used when generating augmented data from the original data, for each virtual speaker

# Speaker	Pitch or speed factor		# Speaker	Pitch or speed factor
1	−2.5	PF	14	0.85	SF
2	−2.0		15	0.9
3	−1.5		16	0.95
4	−1.0		17	1.1
5	−0.5		18	1.15
6	0.5		19	1.2
7	1.0		20	1.25
8	1.5		21	1.3
9	2.0		22	1.35
10	2.5		23	1.4
11	0.7	SF	24	1.45
12	0.75		25	1.5
13	0.8		26	1.55