I've previously written a rendering class that does does just that:

package PDF::ToText;

use 5.006;
use warnings;
use strict;
use CAM::PDF;
use CAM::PDF::GS;
use base qw(CAM::PDF::GS);

=head1 NAME

PDF::ToText - CAM::PDF renderer to extract PDF Text and position infor
+mation

=head1 VERSION

Version 0.01

=cut

our $VERSION = '0.01';

=head1 SYNOPSIS

    use CAM::PDF;
    use PDF::ToText;
    my $pdf = CAM::PDF->new($filename);
    my $contentTree = $pdf->getPageContentTree(1);
    $contentTree->render("PDF::ToText");

=head1 SUBROUTINES/METHODS

=head2 renderText

=cut

sub _textToDevice {
    my $self = shift;

    my @t2u = $self->textToUser( @_ );
    my @t2d = $self->userToDevice( @t2u);

    return @t2d;
}

sub renderText {
   my $self = shift;
   my $string = shift;
   my $width = shift;

   # collect vertices of this text segment.

   my @bottom_left = $self->_textToDevice(0, 0);
   my @bottom_right = $self->_textToDevice($width, 0);
   my @top_left = $self->_textToDevice(0, $self->{Tfs});
   my @top_right = $self->_textToDevice($width, $self->{Tfs});

   printf "%7.2f %7.2f %7.2f %7.2f %s\n", @bottom_left, @top_right, $s
+tring; 

   return;
}
[download]

It's a drop in replacement for CAM::PDF::PageText.

In it's current state, it dumps text coordinates to STDOUT; but it can be easily amended to collect them in a global variable or whatever (CAM::PDF doesn't currently support the passing of handles).