If you want to try, here are the steps you will need to do.

1. You will need to run alsactl (through system or a piped open) to make sure your input is set to Linein (as opposed to Mic). It would go something like this:

#setup a shell thru IPC to carry out different                        
# mixer control commands
my $pidcon = open3(\*CON,0,0,'/bin/sh');
# ...... ..... snip ......
#set Line as input
print CON "amixer sset Line Capture cap\n";
[download]

2. Then you need to record 5 seconds of input:

#!/usr/bin/perl
use Audio::DSP;
#alsamixer must be setup right, just turn everything up :-)

($buf, $chan, $fmt, $rate) = (4096, 1, 16 , 8000);

$dsp = new Audio::DSP(buffer   => $buf,
                     channels => $chan,
                     format   => $fmt,
                     rate     => $rate);

$seconds = 5;
$length  = ($chan * $fmt * $rate * $seconds) / 8;

$dsp->init() || die $dsp->errstr();

open(OUT, ">out.raw") or warn $!;

# Record x seconds of sound
for (my $i = 0; $i < $length; $i += $buf) {
        $dsp->read() || die $dsp->errstr();
        print OUT $dsp->data();    
       }

$dsp->close();
[download]

3. Then once you have that input, you can compare it with a FFT (Fast Fourier Transform) to the 5 seconds of song. See PDL::Audio and Re: Manipulating Audio Data in Perl

Finally, you have about 3 months of work to perfect this. If you get it to work, the Dept. of Homeland Security probably would give you a job in it's voice analysis unit. :-)