There's no need to strip off leading and trailing spaces if you split using a regex with optional whitespace around your delimiters. I get around the need for a %temp hash by using cascading maps.

use strict;
use warnings;

use Data::Dumper;

open my $csvFH, q{<}, \ <<EOD or die $!;
Person1, Name = Lydia, Age = 20, Gender = F
Person2, Name = Carol, Age = 54, Profession = Student, Gender = F, Hei
+ght = 4'8
Person3, Name = Andy, Age = 37, Location = USA, Gender = M, Weight = 1
+17
Person4, Name = Nick, Age = 28, Gender = M
EOD

my %people =
   map {
          $_->[ 0 ],
          { map { split m{\s*=\s*} } split m{\s*,\s*}, $_->[ 1 ] }
       }
   map {
          chomp;
          [ split m{\s*,\s*}, $_, 2 ]
       }
   <$csvFH>;

print Data::Dumper->Dumpxs( [ \ %people ], [ qw{ *people } ] );
[download]

I hope this is of interest.