Voice box supreme: FPGA-based hardware implementation of an ADPCM codec

• Main Authors: Cada, Rona Francesca N., Javier, Joyce Patricia A., Te, Mark Anthony D., Yu, Rhydon William C.
• Format: text
• Language: English
• Published: Animo Repository 2007
• Subjects: Pulse-code modulation; Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/14308
• Institution: De La Salle University

This is an FPGA hardware implementation of the ITU's G.726 Recommendation for Adaptive Differential Pulse Code Modulation (ADPCM). ADPCM is a speech compression technique that can compress speech by as much as 400%. The project makes use of the Virtex-II V2MB1000 development board. Voice is sampled using an analog-to-digital pulse code modulation (PCM) encoder which converts the speech into a 13-bit PCM code. It is then compressed to 32 kbps and stored on an external static random access memory. The group was able to implement the G.726 algorithm on Matlab and VHDL. The Matlab code was designed for comparison with the VHDL output. The group recorded six speech samples from different speakers and tested the outputs for their signal-to-noise ratios. All the samples scored higher than 20 dB. The ACR Subjective Listening Test was also performed on these samples and all outputs scored fair/moderate. In the future this can be applied in Voice over Internet Protocol, speech archiving, and cordless telephony.