Objectives
The aims of this assignment are:
1. To understand and implement predictive coders.
2. To implement speech coding algorithms and gain an understanding of their computational complexity.
Question 1
Write a speech encoder using the LPC algorithm. You must implement your own LPC code. You must not use any MATLAB "toolbox" functions - that is, any which are not shown as toolbox\matlab in response to the which command. Use the following parameters:
1.A frame size of 200 samples for 8kHz sampling rate; Note that if using a sampling rate other than 8kHz, you will need to scale the frame size accordingly.
2.10 th order prediction;
3. Single-pulse excitation at the start of the frame; 4.RMS energy normalization.
Include a listing of the code and an explanation (in your own words) of each stage of the encoding. Comment on the audio quality of the reconstructed speech in terms of intelligi- bility and naturalness using a male speaker and a female speaker. Then encode an audio file containing music and comment on the result.
Question 2
Vary the frame size over what you consider to be a "reasonable" range and comment on the audio quality of the reconstructed speech using male and female speakers. For the optimal frame size found above, vary the number of LPC parameters. Is 10 a reasonable choice?
For each frame size and speaker, tabulate the bit rate in bits per second (bps) required using 8-bit scalar quantization for the RMS energy parameter and each of the ten LPC coefficients. You do not need to quantize the parameters, just assume that 8 bits per parameter is required and calculate using that.
Question 3
Enhance the coder using pitch information, as follows. For each frame, calculate the pitch using autocorrelation. Then generate the excitation using impulses spaced at the pitch interval. Test using male and female speakers, and comment on any improvements you notice over single-pulse excitation. Note that you may need to buffer one or more previous frames in order to calculate the correlation, and that the placement of the pitch impulses in the current frame depends on the location of the last pitch pulse in the previous frame.
Include a listing of the code and an explanation (in your own words) of each stage of the encoding.
Comment on the perceived quality of the LPC coder with pitch excitation as compared to single-pulse excitation.