Re: Fw: sursound: speaker phones (Henrik =?iso-8859-1?Q?M=F8ller?= )

Subject: Re: Fw: sursound: speaker phones
From:    Henrik =?iso-8859-1?Q?M=F8ller?=  <hm(at)KOM.AUC.DK>
Date:    Wed, 4 Jul 2001 10:57:51 +0200

This is a multi-part message in MIME format. --------------329139244EE953A7CB91E14C Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Dear James (and list). I just came across your old question below, which - as far as I can see - was only partly answered in the following postings. The full answer is probably not available now, but I have a few more comments that may be of help. I put my comments into your text. "James W. Beauchamp" wrote: > While we're on the subject of sound localization, can someone explain > why speaker phones always sound like you're "talking through a tube" to > the person on the other end of the line? I'm radiating a sound which > is picked up by a diaphragm on a table and then directly transmitted > to someone's ear via a small speaker. How is this substantially > different from my talking to a hole in the table with someone's ear > directly underneath? As other replies also state, it's due to the cocktail party effect. Or to be precise, the reflections (which give rise to coloration or the perception of "talking through a tube") ARE there in both cases (if you listen directly and when you record). BUT, when you are there in person, your cocktail party processor "removes" the reflections, and you don't hear them. When you listen to the recording, the cocktail processor does not work, since you do not have the proper binaural signals. (We have a setup, in which you listen through a "hole in a wall" (with one ear). It sounds quite strange). > > Here is a related problem: Suppose I wish to record someone talking in > the front of the room, and I am in the back of the room. When I am > actually there listening, the speech is as clear as a bell; I ignore > all environmental sounds and echoes. To (roughly) simulate the pressures > at the ears, I take the headphone of my Walkman, put it on, and use it > as a stereo microphone. Later, when I play it back through the headphones, > the basic sound is there, but now the echos and environmental sounds swamp > out the speaker, who is rendered barely audible. Does using really good > mics help? (Cheap actual mics don't seem to improve the situation.) This is exactly the same. In the back of the room there is probably much more reverberant than there is direct sound. So your cocktail party processor is really working and doing a good job in real life. This does not happen with the recording. The fact that you use a stereo recording does not help. Most stereo microphones are two directional microphones put at an angle, so that the intensity of a given sound is different in the two channels, but there are no interaural time differences. Such a recording can give an excellent stereo perspective, when it is played back through loudspeakers. This is not the place to go into details about how and why this works, but it is important to be aware that the sound from the two channels must be combined as they are in the sound field from the two loudspeakers. When you listen through headphones this does not happen, and you get the perception of a sound field inside your head without clear directions to individual sound sources. If you use a binaural recording (with a good dummy head or yourself, and properly equalized), the cocktail party effect works again. It is also my experience that it works fairly well, if you use a stereo recording, where you introduce the proper interaural time differences. You may approximate this quite well by using two omnidirectional microphones set at a distance of, say, 17 cm, thus simulating the two ears. This technique approximates a binaural recording; it does give some degree of - if not full - externalization, but of course it does not give any perception of back or front, or up or down. (A combination - two spaced AND angled directional microphones - is sometimes used for stereophonic recordings meant for loudspeaker reproduction). > > If we understand what the problem is, how do we correct for it? E. g., > why aren't there better speaker phones? (Maybe there are for a price.) > > I realize that this problem is being worked on in the context of hearing > aids -- my neighbor has to take his off in order to hear a conversation > when there's more than 1 other person talking -- and in tele-conferencing > applications. > > Jim Beauchamp > Univ. of Ill. U-C So, don't spend too much money on more expensive microphones, but use the cheap ones properly. Or buy a dummy head (which is not cheap, unfortunately). Sincerely, Henrik Møller --------------329139244EE953A7CB91E14C Content-Type: text/x-vcard; charset=us-ascii; name="hm.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Henrik Møller Content-Disposition: attachment; filename="hm.vcf" begin:vcard n:Møller;Henrik tel;fax:+45 9815 2144 tel;work:+45 9635 8710 x-mozilla-html:TRUE url: org:Aalborg University;Department of Acoustics adr:;;Fredrik Bajers Vej 7 B4;Aalborg Ø;;DK-9220;Denmark version:2.1 email;internet:hm(at) title:Professor, Head of Department fn:Henrik Møller end:vcard --------------329139244EE953A7CB91E14C--

This message came from the mail archive
maintained by:
DAn Ellis <>
Electrical Engineering Dept., Columbia University