Code excited linear predictive speech coding implementation on DSP

• Main Authors: Anunciacion, Amiel, Casquero, Mark Anthony, Ching, Carolyn Grace, Onan, Jaquelyn
• Format: text
• Language: English
• Published: Animo Repository 2000
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/11882
• Institution: De La Salle University

CELP or Code Excited Linear Prediction is capable of coding a relatively good quality speech at low bit rates, mainly having 4800 bps, for the FS1016 Standard. Since CELP algorithm involves complex mathematical computations, familiarization of different theories is done through MATLAB simulations. This thesis incorporates the CELP algorithm into two DSP processors. The Matlab code is translated to C code which is compiled into the TMS320C54X speech into 144 bits per frame. The encoded bits are transmitted to the other EVM via connecting wires. The bits received by the second EVM are decoded and processed into a synthesized speech. After processing, the final decompressed speech values are monitored using the C source debugger. These values are manually obtained and are made play in Matlab. The main purpose of this thesis is to implement the CELP system, with encoder and decoder onto two DSP Evaluation Modules. The efficiency of the system/quality of output speech is determined quantitatively by obtaining the signal to quantizing noise ratio of the input versus the output speech. the result is audible speech. The analyzer part of the CELP system is not implemented on the DSP chip due to the processor's fixed point and program complexity limitations. Moreover, module interface failed because of technical problems in synchronization. However, manual processing is successful since the sample by sample speech values obtained from the debugger is played audibly when transferred to Matlab. The main purpose of this thesis is to implement the CELP system, with encoder and decoder onto two DSP Evaluation Modules. The efficiency of the system/quality of output speech is determined quantitatively by obtaining the signal to quantizing noise ratio of the input versus the output speech. The result is audible speech. The analyzer part of the CELP system is not implemented on the DSP chip due to the processor's fixed point and program complexity limitations. Moreover, module interface failed because of technical problems in synchronization. However, manual processing is successful since the sample by sample speech values obtained from the debugger is played audibly when transferred to Matlab.