Speech Processing by Computer

 

LAB 1

EDITING OF SPEECH SIGNALS

 

This lab session demonstrates the effect of choice of sampling rate and quantisation on the quality of a digitised speech signal.  The effects of 'aliasing' distortion arising from sampling at too low a rate are heard.  The use of SFS programs for editing signals is explored.

 

1.         Choice of sampling rate

            (i)         Acquire a speech signal at 20000 samples/second.

            (ii)        Use the Tools/Speech/Process/Resample command to resample the signal at 16000, then 12000, then 8000 samples per second.

            (iii)       Listen to the four versions.  How do they differ?  Why?

 

2.         Choice of quantisation level

            (i)         Acquire a speech signal at 20000 samples/second.

            (ii)        Run the program 'qchange' and enter the name of your speech file.

            (iii)       Select the number of bits of quantisation (from 16 downto 1) and listen to the replayed version.  What happens?  Why?

 

3.         Aliasing distortion

            (i)         Acquire a speech signal at 20000 samples/second.

            (ii)        Run the program 'rchange' and enter the name of your speech file.

            (iii)       Select a new sampling rate between 6000 and 20000 and listen to the replayed version.  What happens?  Why?  What is the difference between what 'rchange' does and proper resampling?

 

4.         Editing

            (i)         Acquire into separate files the utterances:

                                    "The time is"                 into file 't1.sfs'

                                    "ten"                             into file 't2.sfs'

                                    "minutes past"               into file 't3.sfs'

                                    "twelve"                        into file 't4.sfs'

            (ii)        Using the Tools/Speech/Edit command, select annotation mode with annotation labels of type ‘marker’. Add markers into the files indicating where the speech signal starts and stops in each file.  Use the markers 'start' and 'stop'.

            (iii)       Using the ‘notepad’ editor, create a text file 'time.txt' containing:

                                    t1.sfs,,,start,stop

                                    t2.sfs,,,start,stop

                                    t3.sfs,,,start,stop

                                    t4.sfs,,,start,stop

            (iv)       Create a new empty SFS file

            (v)        Run the program Tools/Generate/Concatenate signals to edit the component parts together

(vi)              Replay the concatenated signal.  What are the problems with your concatenated utterance?  How might it be improved? 

            (vii)      Change your files to make the utterance 'The time is twelve minutes past ten'.