Cite this article:
Khan (2007). Reduced Complexity Low Delay, High Quality Spectrally Efficient Algebra Code Excited Linear Prediction (Acelp) Speech Coding Algorithm. Journal of Ravishankar University (Part-B: Science), 20(1), pp.1-12.
Journal of Ravishankar University Vol. 20 No. B (Science) 2007 PP 1-12 ISSN 0970-5910
Reduced Complexity
Low Delay, High Quality Spectrally Efficient Algebra Code Excited Linear
Prediction (Acelp) Speech Coding Algorithm
M.R. Khan
NIT Raipur, C.G., India
mrkhan63@gmail.com
Abstract:
this paper describes the implementation of spectrally efficient ACELP based
speech coding algorithm with a target bit rate of 7.2 per second (Kbps)
suitable for speech communication over or wired line Communications. The 8 kbps
ITU G.729 [1] best speech coding algorithm has been modified so as to achieve
lower bit rate and targeting low computational and search complexity, at the
same time maintaining nearly the similar quality of speech reconstruction. The
short-term spectral estimates are updated every 10ms (frame) and encoded as
line spectrum frequencies (LSF). In order to reduce the bit rate and search
complexity a pre-selection based sub-optimal, intraframe LSF vector
quantization is proposed. The long-term protection (LTP) parameters are updated
every 5ms (subframe) and implemented as modified open loop method (MOLP) with a
pre-selection over a tri-valued speed signal, which considerably reduces the
search complexity of the adaptive code book search. An algebraic structure and
fixed code book based excite ation is proposed with reduced search complexity
and combined optimisation of the LTP gain and fixed codebook gain for every
subframe. The simulation results show that the performance of the proposed
speech coding algorithm is nearly same as G.729 with lesser complexity and
shorter processing delays.
Keywords: Speech coding, ACELP
NOTE: Full version of this manuscript is available in PDF.