It looks like this encoding method is very similar to the BER compressed integer format except that 0 instead of 1 is used to signal continuation into the next byte. So the following should work:

sub byte_align_decode {
  my $bytes = shift;
  $bytes ^= "\x80" x length($bytes);
  unpack("w*", $bytes);
}

# from the example in the above URL

print join(' ', byte_align_decode("\x06\xb8\x85\x0d\x0c\xb1")), "\n";
# emits: 824 5 214577
[download]

The encoding routine is similar:

sub byte_align_encode {
  my $bytes = pack("w*", @_);
  $bytes ^= "\x80" x length($bytes);
}
[download]