The unit is powered with a 5v wall adapter and alligator clips instead of four 1.5V coin cells. The mode switch (at the lower right) is set to "OPER" and a letter 'P' is shown in the screen. Enter "0406091001" and press "ent"; a letter 'H' is now shown. Enter "001290" and "ent", a '0' is displayed: that was the end user ID. Now, different product codes can be stored.

Begin with 'H' or 'P' and type digits, then press "ent". Information is briefly displayed like:

0004      0451

with the first number increasing and the second decreasing. Probably means the number of occupied and free memory positions. After a few seconds, '0' can be seen again. The arrow keys let you check previously stored data but I don't know how to edit or delete a position.

After storing a few codes as "H12345678" or "H96385274" I place the mode switch in "SEND" and press button "SND/=". The screen displays "SEND BUSY" and this audio sequence is emitted:

For 4.4s a 1.200 Hz tone is heard. Then, for 6s, it is 1 kHz. For 2.8s there is a modulation. For the last 3s, the 1 kHz tone is sent again. A careful look at the spectrum shows the frequencies are not very accurate. The tones actually seem 980 and 1180 Hz.

I installed this program: http://www.whence.com/minimodem/minimodem.1.html

and ran:

minimodem -r -f msi55_bell103_3.wav -M 980 -S 1180 300

which returned:

### CARRIER 300 @ 1000.0 Hz ###
�H00��90+�H00��90+�H00��90+�H��3�56��+�Ʊ�3�56��+��9��+�ƴ56+�H963�5���+�
### NOCARRIER ndata=74 confidence=2.026 ampl=0.147 bps=294.55 (1.8% slow) ###

This looks like a Bell 103 modulation: https://en.wikipedia.org/wiki/Bell_103_modem

Even though Bell 103 has 1070 and 1270 Hz tones instead of 1000 and 1200 as I found.

Maybe the components have aged and the tone and symbol frequency have drifted. I edited the audio clip for speeding it up 1.8%. The audio tones were now close to 1000 and 1200 Hz and the symbol rate close to the theoretical 300 bps:

minimodem -r -f msi55_bell103_4.wav -M 1000 -S 1200 300 -R 8000 -8 --startbits 1 --stopbits 1

### CARRIER 300 @ 1000.0 Hz ###
�H00��90+�H00��90+�H00��90+�H��3�56��+�Ʊ�3�56��+��9��+�ƴ56+�H963�5���+�
### NOCARRIER ndata=74 confidence=2.090 ampl=0.148 bps=299.50 (0.2% slow) ###

In both cases the demodulated data makes sense in spite of some errors. The code "H12345678" is now "H��3�56��". It could be due to some 50 Hz mains noise on top of the modulation, caused by the cheap wall adapter. minimodem program reports a low confidence value (2), consisten with a distorted signal. All in all, the esential features and technical specs have been understood.