Self-Excited Vocoder
Speech-Compression
Algorithm
by Atlanta Signal Processors, Inc.

Software Overview

The Self-Excited Vocoder (SEV) is a high-quality, medium-to-low bit-rate coder algorithm. It is a 
member of the class of speech coders, based on the analysis-by-synthesis coding method, that offers 
the highest speech quality at data rates in the range of 4.8 kbps to 9.6 kbps. A distinguishing 
characteristic of the SEV is that it produces speech that sounds more natural, more intelligible, 
and less mechanical than is typical for speech coders operating at similar data rates.
The SEV, operating at 9.6 kbps, offers speech quality comparable to waveform coders operating at 16 
kbps. However, as with all parametric coders of this type, this high speech quality at low 
bit-rates comes at the cost of relatively high computational complexity. The full-duplex SEV 
implementation consumes 95% of the processing power of a Texas Instruments TMS320C30 processor 
operating at 33 MHz. The speech analysis section of the algorithm is far more computationally 
demanding than the synthesis section; the analysis accounts for 91% of the CPU cycles of a 33-MHz 
`C30.
The SEV algorithm is optimally configured for a 9.6-kbps data rate. The quality of the speech 
produced by the SEV at 9.6 kbps is equivalent to the quality produced by the Subband Coder at 16 
kbps. The SEV can be configured to provide data rates as low as 4.8 kbps. The quality of the coded 
speech degrades proportionately as the data rate is reduced.
Selecting and implementing analysis-by-synthesis speech coding methods in the SEV is the direct 
result of research and development backed by over 25 years of experience in the science of speech 
compression. The SEV's coding and quantization methods, the result of painstaking research and 
experimentation, have been optimized to yield an extraordinary level of fidelity and resistance to 
corruption in the speech signal and quantized parameters.  

Features and Benefits

•	High speech quality at medium bit rates

Processor and System Specifications

•	Texas Instruments TMS320C3x, operating at 33 MHz or above.
•	Supporting memory: 128 k ¥ 32-bit off-chip RAM
•	Signal input: 16-bit linear
•	33-MHz `C30 utilization: 80% at 4.8 kbps, rising to over 95% at 9.6 kbps

Usage Limitations or Performance Considerations

•	The encoder is CPU intensive.

Algorithm Verification

•	There are no international standards for this algorithm. However it has been extensively tested.

Availability

•	The Self-Excited Vocoder algorithm is available now from Atlanta Signal Processors, Inc.

Company Background and Contact Information

Atlanta Signal Processors was founded in 1981 by Ronald W. Schafer, Thomas P. Barnwell III, and 
Russell M. Mersereau, three Professors of Electrical Engineering at the Georgia Institute of 
Technology with over 80 years of experience in the field. The company has focused its efforts on 
developing tools for professional signal-processing design engineers and now offers digital 
signal-processing hardware and software products to DSP designers worldwide.
Contact:	Atlanta Signal Processors, Inc.
		1375 Peachtree St. NE, Suite 690
		Atlanta, Georgia  30309-3115
		(404) 892-7265
		Fax: (404) 892-2512
		e-mail: aspi@aspi.com