A 5‐kHz‐bandwidth low‐bit‐rate speech coder: Bandwidth, noisiness, and opinion scores
Noboru Harada, Hitoshi Ohmuro, Kazunori Mano · Systems and Computers in Japan · 2002
Abstract With the development of PCs, Internet technology, and mobile communications, the demand for low‐bit‐rate speech coding schemes as well as for high‐information‐level natural speech have increased in recent years. While naturalness increases with the use of speech signals of broad frequency bands, more information is needed for coding. In this paper, we propose a 5‐kHz‐band speech coding scheme that clarifies the relationships between the frequency band feeling, noisiness, and subjective opinions on quality by means of evaluation experiments using multiple frequency bands, while allowing selection between low bit rate and naturalness. The proposed scheme operates in three bit‐rate modes, 7.8, 5.75, and 3.95 kbit/s. Subjective quality evaluation tests confirm that the coding scheme achieves quality almost equal to G.729 (3.4‐kHz bandwidth, 8 kbit/s) and G.722 (7‐kHz bandwidth, 48 kbit/s) at 5.75 kbit/s and quality higher than these at 7.8 kbit/s. © 2002 Wiley Periodicals, Inc. Syst Comp Jpn, 33(14): 110–119, 2002; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/scj.1181