Real-Time Implementation Of LPC-10 Codec On TMS320C6713 DSP
During last two decades various speech coding algorithms have been developed. The range of toll speech frequency is from 300 Hz- 3400 Hz. Generally, human speech signal could be classified as non-stationary signal because of its fluctuation randomly over the time axis. One important assumption made...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2012
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/24122/1/Real-Time%20Implementation%20Of%20LPC-10%20Codec%20On%20TMS320C6713%20DSP%20-%20Wissam%20Talib%20Alshammari%20-%2024%20Pages.pdf http://eprints.utem.edu.my/id/eprint/24122/2/Real-Time%20Implementation%20Of%20LPC-10%20Codec%20On%20TMS320C6713%20DSP.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | During last two decades various speech coding algorithms have been developed. The range of toll speech frequency is from 300 Hz- 3400 Hz. Generally, human speech signal could be classified as non-stationary signal because of its fluctuation randomly over the time axis. One important assumption made to make the analysis of such signal even easier by assuming the speech signal is quasi-stationary over short range (frame). The frames of speech signal can be classified further into Voiced or Unvoiced, where the
voiced part is quasi-stationary while the unvoiced part as an AWGN. The quality of the synthesized signal is degraded significantly due to the excitation of voiced part not equally spaced within the frame and the excitation of the unvoiced part is not exact AWGN. This assumption produced a non-natural speech signal but with high intelligible level. One more reason is that the frame could have voiced plus unvoiced parts within the same frame, and by classifying this frame as voiced or unvoiced due to rigid decision would drop the level of quality significantly. Speech compression commonly referred to as speech coding, where the amount of redundancies is reduced, and represent the speech signal by set of parameters in order to have very low bit rates. One of these speech coding algorithms is linear predictive coding (LPC-10). This thesis implements LPC-10 analysis and synthesis using Matlab and C coding.
LPC-10 have been compared with some other speech compression algorithms like pulse code modulation (PCM), differential pulse code modulation (DPCM), and code excited linear prediction coding (CELP), in term of segmental signal to quantization noise ratio SEG-SQNR and mean squared error MSE using Matlab simulation. The focus on LPC-10 was implemented on the DSP board TMS320C6713 to test the LPC-10 algorithm in realtime. Real-time implementation on TMS320C6713 DSP board required to convert the Matlab script into C code on the DSP Board. Upon successfully completion, comparison of the results using TMS320C6713 DSP against the simulated results using Matlab in both
graphical and tabular forms were made. |
|---|