a good descriptive guide book with matlab coding for neural networkDescripción completa
Speaker Recognition is a process of automatically recognizing who is speaking on the basis of the individual information included in speech waves. Speaker Recognition is one of the most useful biom...
image processing using matlabFull description
a good descriptive guide book with matlab coding for neural network
QUANTIZATION AND SAMPLING USING MATLAB Introduction In this part of the voice communication lab a simulation tool (Matlab) is used to experiment with quantizing and sampling audio. The idea is you get to change sampling and quantization parameters and hear the effects. For background see lecture slides and supporting notes. This part of the lab needs to be done in CIB 215 as Matlab software is required. The Matlab environment allows us to experiment with processes like quantization and sampling in software as if it was being done in real hardware. The lab can be undertaken in pairs although you are encouraged to discuss results with the rest of your group. The lab Copy model files (*.mdl) from NOW (.zip file) into MATLAB directory in your documents folder. Run Matlab (start menu => programs)
Communication Technology
RG '11
First record a sample file using the record model (double-click the model name) there is also some sample music (music 1.avi and music 2.avi). If the files aren’t listed make sure you have the correct directory. Run the simulation by pressing the black arrow key. The default is to run for 10 seconds (you can make a longer sample by changing the 10 to the right of the arrow to a bigger value). The audio sample will be stored in a file called 'input.avi'. You can change this by double-clicking the 'to Multimedia file' box. To change the audio sample at any time simply re-run the record model.
Now you have some audio to use as input you are ready to experiment with quantization and sampling. At this stage it should be possible to test the effects of quantization and sampling in real-time. That is run the models and send the output directly to the audio output (=headphone socket). However PCs in CIB215 appear too slow so an interim step of sending the output to another avi file is required.
Communication Technology
RG '11
Quantization Open quantization.mdl. Run the model. Then use play.mdl to listen to the result (use the headphones). Note any differences with the original unquantized 1 audio. Now run the quantization model again but this time changing the quantization step size first (double-click on the quantizer and increase the quantization-step parameter). Note any effects on the quantized audio. Repeat this exercise with progressively larger values of quantization step. At what step-size is the quantization just noticeable? Sampling Now open and run the sampling model. The default is to not change the sample rate so everything should sound fine. Now run the model again this time changing the downsample, upsample and gain parameters to 2. Note any effects on the resulting audio. Repeat increasing the parameters by 1 each time. About the model: To change the sampling rate of the audio we throw away some of the samples (downsample block). So that we can listen to the result we need to put the sampling rate back to the original rate (44.1kHz) as required by the audio output. To do this the downsampled signal is first filtered then upsampled by the same factor as the downsample. This then needs to be filtered and amplified (again by the same factor). The result is a signal which behaves as if it was sampled at 44.1kHz/downsample-factor. Repeat the exercises with the sample audio in sample1.avi. Are your results different? Conclusion Include this exercise in the report for the overall lab (about one side), record and explain your results.
1 Note: The original audio sample is already sampled and quantized. However the sample rate (44.1kHz and 16 bits) do not result in noticeable affects such as aliasing and noise. Quantization.mdl and sampling.mdl allow us to change the quantization step and sampling rate respectively.