Microphone to pick up voice recognition in large room...will array work?
DD
speech commands in a 20x12 room wil 8' ceiling? I am working
on a home automation project to tie voice recognition. Using
Dragon Dictate or IBM via voice to run marcros to control X-10
units.
I have heard of array microphones, but they seem to have a
limit of 36" distance range. This microphone solution would
have to pick up sound in any part of the room.
Any help/ideas/advice is appreciated.
H...@xxxhydrologistxxx.com
<g0ifrtcgh2v8cbn3s...@4ax.com>:
What you want to do is not practical in any but specialized anechoic
acoustic environments. An microphone array won't solve the problem.
See:
http://shure.custhelp.com/cgi-bin/shure.cfg/php/enduser/std_adp.php?p_sid=bBnhnwYf&p_lva=&p_refno=000217
-000014&p_created=950774400&p_sp=cF9ncmlkc29ydD0mcF9yb3dfY250PTE4NDcmcF9wYWdlPTM*&p_li=
Also:
http://shure.custhelp.com/cgi-bin/shure.cfg/php/enduser/std_adp.php?p_sid=bBnhnwYf&p_lva=&p_refno=
000126-000004&p_created=948873600&p_sp=cF9ncmlkc29ydD0mcF9yb3dfY250PTE4NDcmcF9wYWdlPTU*&p_li=
If you really still want to pursue sound control:
1) plan on being close to the microphone
2) use a gated mixer so that only one mic is "open" during recognition
3) don't expect perfection.
HTH ... Marc
Marc_F_Hult
H...@xxxhydrologistxxx.com
Lewis Gardner
>limit of 36" distance range. This microphone solution would
>have to pick up sound in any part of the room.
From my sound reinforcement days...
In general microphones need to be placed three times the working distance
from each other to be effective. I was told this had something to do with
phase cancellation and was a rule of thumb and not an absolute. This means
that if a microphone is 10 feet from the person speaking the microphones
need to be 30 feet apart. This assumes all the mics on all the time. In a
whole room situation it will result in "dead" zones.
You might want to look into PZM type microphones and "automatic" mixers.
Anthony Matonak
>
> Does anyone know of a microphone solution that will help pick up
> speech commands in a 20x12 room wil 8' ceiling? I am working
I have no solution but some passing thoughts. As I see it, the
main problem with more than one microphone covering the same
area will be that they will pick up the same sound at slightly
different times. You couldn't just mix them together directly
or you'll get all kinds of weird artifacts, echoes, phase
cancellation and the like. You can't just use a fixed 'time delay'
like they use on stadium speakers because you won't know where
the speaker is going to be speaking from and they could even be
moving.
I can think of only two ways to make this work. First way is to
run a separate voice recognition program for each single microphone
and then have a program mix/check/compare the outputs of them all
and have it discard the garbled recognition and duplicates. I'm
not sure how your recognition software operates but most sound
cards have at least two inputs (left & right) and sometimes you can
put as much as two or three sound cards into a single machine.
You could potentially run as much as 4 or 6 microphones off one
CPU. Maybe more with the right kind of sound cards or some kind of
microphone to USB adapter. Alternatively, you could simply run a
bank of cheap computers, single board PC's or dedicated signal
processors with a one CPU per microphone arrangement. I've seen
people rig up banks of motherboards that have everything built in,
video & network card especially, all set to do network booting.
The other alternative is to have some kind of dynamic time correcting
mixer. You would plug all your microphones into it and it would compare
the sound arriving at each and adjust the time delay dynamically as it
mixes all their outputs together. The hard part with this is that if
you have more than one speaker or source of sound at a single instant
it will need to be able to separate them and time shift each sound
differently. It might be able to do this using a spectral analysis.
On the up side, if it CAN do this then with a map of the 'known'
positions of the microphones it could probably also plot the relative
location of each source of sound in the room much like a submarine sonar.
If you place a few microphones at floor level as well as ceiling it would
even be able to place the sources in a 3D space. You could then make your
commands do different things relative to the place or height they were
spoken from. By designating the radio and TV as places where commands
will be ignored you could eliminate the evil 'clapper syndrome' where loud
gunfights on TV would turn the lights on and off. In fact, you could go
one step further and have things happen simply based on any sound coming
from a specific location. This could be refined to commands like 'knock
three times on the ceiling' or 'twice on the pipe'.
Probably cheaper to go to a spy shop and get one of those fancy 'pick
up a whisper at 30 feet' microphones and just use one for the entire
room.
Anthony
Anand Dhuru
pocket all time? Feed it thru' a lapel mike. An FM receiver close to
the PC then outputs the commands thru' the line-in on your sound card.
Regards,
Anand Dhuru
Andrew Ward
>
> On 30 Sep 2001 20:39:17 -0500, DD <don...@mindspring.com> wrote in message
> <g0ifrtcgh2v8cbn3s...@4ax.com>:
>
> >Does anyone know of a microphone solution that will help pick up
> >speech commands in a 20x12 room wil 8' ceiling? I am working
> >on a home automation project to tie voice recognition. Using
> >Dragon Dictate or IBM via voice to run marcros to control X-10
> >units.
> >
> >I have heard of array microphones, but they seem to have a
> >limit of 36" distance range. This microphone solution would
> >have to pick up sound in any part of the room.
> >
> >Any help/ideas/advice is appreciated.
>
> What you want to do is not practical in any but specialized anechoic
> acoustic environments. An microphone array won't solve the problem.
BZZZZT
I'm doing it. No problem.
15x20 ft room, 8 ft ceiling
Crown PZM-11 mounted in the ceiling near the center.
Shure SCM810 Automatic Mic mixer (5 mics total in the house,
all on simultaneously.)
Works fine. HomeVoice has the most accurate, reliable recognition, in
my experience. I have HAL2000, but have never gotten the accuracy and
reliability as solid as HomeVoice. I am SLOWLY switching over th Hal2000
due to its feature set, but HV is ROCK SOLID reliable and consistent.
take care
-Andrew
> See:
>
> http://shure.custhelp.com/cgi-bin/shure.cfg/php/enduser/std_adp.php?p_sid=bBnhnwYf&p_lva=&p_refno=000217
> -000014&p_created=950774400&p_sp=cF9ncmlkc29ydD0mcF9yb3dfY250PTE4NDcmcF9wYWdlPTM*&p_li=
>
> Also:
> http://shure.custhelp.com/cgi-bin/shure.cfg/php/enduser/std_adp.php?p_sid=bBnhnwYf&p_lva=&p_refno=
> 000126-000004&p_created=948873600&p_sp=cF9ncmlkc29ydD0mcF9yb3dfY250PTE4NDcmcF9wYWdlPTU*&p_li=
>
> and
> http://shure.custhelp.com/cgi-bin/shure.cfg/php/enduser/std_adp.php?p_refno=000225-000002&p_created=951465600
>
> If you really still want to pursue sound control:
>
> 1) plan on being close to the microphone
> 2) use a gated mixer so that only one mic is "open" during recognition
> 3) don't expect perfection.
>
> HTH ... Marc
> Marc_F_Hult
> H...@xxxhydrologistxxx.com
--
-Andrew
------------------------------------------------------------
Andrew Ward
ph: 408 527-2773 | |
800.250.4800 ext 72773 | |
Public Carrier IP BU ||| |||
GSR development group (12000) ..:|||||||:...:|||||||:..
email to: aw...@cisco.com C i s c o S y s t e m s
------------------------------------------------------------
H...@xxxhydrologistxxx.com
message <3BC4D864...@cisco.com>:
Hi Andrew:
Your recommendation of a pzm/boundary-type mic is an excellent one, Shure's
admonitions in the url I cited notwithstanding (' Microphones On The
Ceiling? Shure Strongly Advises "NO!"' and " If you decide to install
ceiling microphones ... we would prefer that you not use Shure
microphones". I note that you use Crown microphones ;-)
It is also useful to explicitly point out to the original poster (who plans
to use dictation software) that many folks have had better experiences with
voice control software designed for Home Automation (HAL, HomeVoice) than
with dictation software (Dragon, ViaVoice). Any experience out there with
Windows XP?
But your "no problem" assertion seems inconsistent with your previous post
to this newsgroup to the effect that you did have problems, including when
you had a cold. (Sorry, Honey. You'll have to get well before you can watch
TV again ...)
And the $2000+ you've spent on just the HomeVoice setup might be a
"problem" for some folks too...
Could you expand a bit on your experience with HomeVoice? Your statement
of "no problem" contrasts with my conclusion of "not practical" with
respect original poster's requirement of voice recognition from "anywhere
in the room".
Do you mean that your HomeVoice recognition setup responds correctly each
time you issue a command in a normal tone of voice wherever you are in a
20' room with a single mic despite simultaneous normal conversation, TV,
music, and so on ? If not, your definition of "no problem" and my
definition of "not practical" overlap ... How do you use voice control to
reduce voice-like ambient sound if you have to have the ambient sound
reduced before you can use voice control (without shouting) ?
The ~$1500 Shure mixer you are using automatically selects the input so
that only one input at a time is presented to the A/D converter. This is
what I meant by gated mixer. I have a pair of surplus Ivie 884's ganged
together to create a gated 16x8 matrix. The Ivie 884x mixers differ from
the Shure SCMx10/FP410 mixers in that although all can gate/filibuster and
have auto level control, some noise reduction needs to be delegated to
microphones' built-in electronics or external compressor/DSP devices rather
than in the mixer itself as with the Shure mixer that you use. Using
outboard compression seems to be HomeVoice's hardware approach too (they
sell a Rolls gate/compressor).
See http://www.dandugan.com/matrix.html for a comparison of many different
automatic mixers.
A nice feature of the Ivie 884's is that they can be controlled both with
remote 10k pots and RS-485 and so can easily controlled with a local
motorized potentiometer or communicate via ASCII commands with homebrew
programs -- or for those other folks whose budget allows,
Crestron/Panja-AMX. Alas, ABIK even with a 70MIPS processor in AMX's latest
extensive, megabuck offerings, AMX has apparently not yet discovered that
voice control throughout a 20'x12' room is "no problem" because, ABIK, they
don't offer _any_ room-wide voice command capability. It will be
interesting to see what they eventually offer (if they don't already).
It has been some time since I last tried HomeVoice, but my past room-scale
tests were consistent with recent experiences with Via Voice and Dragon
using various head sets, a microphone array, and several boundary mics. I
found that even when the volume of extraneous voice/music in the room is
perceived to be "low", I do not get satisfactory results. There are too
many errors and the frequent need repeat commands is awkward and unnatural
(to me).
At more than a 3 feet from an omni microphone and (or) with other
voices/speech/music in the room, performance is entirely unacceptable to
me. Of course if I stand in the middle of a room with 8' ceiling with a mic
on it, my mouth is less than 3 feet from the mic, so that is not a test of
the original poster's needs of "anywhere in the room" (and I don't have any
8' ceilings in my home in any case). I note that in a previous post you
indicated that your test distance was 5 feet which does not correspond to
"anywhere in [a 20'x 12'] room".
I'd be thrilled to find that HomeVoice has the solution, but I suspect not
(for me). If they had a demo of 2.1 that ran on W2k, I'd give it another
try.
Our Mileages Apparently Vary ... Marc
Marc_F_Hult
H...@xxxhydrologistxxx.com
Jack
homebrew solutions? I would like to use some kind of VRecog software to
enhance my home automation setup, and not really tied to any particular
software or hardware as long as it doesn't mean ditching my existing setup
for a new home automation package.
ideas?
Jack
don marquardt
DD
Also, is there any soundcard/solution out there that will let the
computer know *where* the sound came from?
For example if I say "all lights out" in the bedroom,
I don't want all lights to go out in the livingroom also.
DD
H...@xxxhydrologistxxx.com
<14hbtt83bmjs3b2aj...@4ax.com>:
>What is a gated mixer?
>
>Also, is there any soundcard/solution out there that will let the
>computer know *where* the sound came from?
>
>For example if I say "all lights out" in the bedroom,
>I don't want all lights to go out in the livingroom also.
>
Audio mixers electronically combine (add) two or more audio signals. As the
number of input signals increases, so too does resulting overall (hum,
hiss, etc) and ambient (HVAC system, microphone-handling etc) noise level.
In order to minimize noise from these sources, inputs can be automatically
"gated" (opened or closed) so that only those inputs above a determined
threshold contribute to the overall mix. A common use is for multiple
microphones in a conference. In this case, inputs may operate under
"filibuster" mode so that only one input is active at a time and the active
input remains gated open until the speaker (finally!) stops talking.
Some gated (automatic) mixers can electrically indicate which inputs are
active. The Ivie 884's www.ivie.com I am experimenting with have both LEDs
and open-collector transistor outputs that can be used to control a relay
(for example) based on which input is active or can be sensed by a computer
control program. This could be used to instruct the program what/where the
source of the sound is and so accomplish what you want to do.
See http://www.dandugan.com/matrix.html for a comparison of many different
automatic mixers. The price of DSP-based components is dropping so older,
analog versions of mixers, compressors and other audio paraphernalia
useful in building a distributed HA voice-command system are becoming
reasonably priced on the used market. Check out eBay.
H...@xxxhydrologistxxx.com
message <0blbttoqvcj3mgrb3...@4ax.com>:
>The price of DSP-based components is dropping so older,
>analog versions of mixers, compressors and other audio paraphernalia
>useful in building a distributed HA voice-command system are becoming
>reasonably priced on the used market. Check out eBay.
Can y'all keep (or use) a secret?
Of particular interest to HA-DIY'ers are the specialized type of gating
mixers used for teleconference because they facilitate adding telephone
control as well as microphone control. Telephone control is usually
significantly more dependable than room microphone control
Used teleconferencing mixers are available with some regularity on eBay for
a few pennies on the dollar if you know what to look for.
Here's a multi-channel Shure ST6000 Type 2 Teleconference System that I
bought a couple of weeks ago for $24 with multichannel gating, muting,
built-in PA amps, voice-optimized frequency response, volume, gate control
out, etc, etc
http://cgi.ebay.com/aw-cgi/eBayISAPI.dll?ViewItem&item=1643047959
http://www.shure.com/pdf/discontinued/ST6000T2_installers_manual.pdf
Rick Tinker
All good points regarding the differences.
I am coming into this quite late as I don't monitor UseNet News quite
like I used to, but...
You can have all of the microphones connected to a mixer on at the
same time with a system like HomeVoice because it is speaker
dependent. As Andrew points out, it works great with all of them on,
however I have posted many of the downsides of this system in this
newsgroup before and as Mr. Hult points out, changes in your voice
causing problems is very real with this type of system.
If you do go with a speaker independent system like HAL2000, then you
do need to have gating and a filibuster mode. It's not a limitation
of the ASR technology, it's just because of the difference that
speaker independent ASR is processing all sounds for recognized words
spoken by anybody and speaker dependant systems can pull the exactly
matched spoken phrase from a stream of audio sounds.
The Shure SCM410 and SCM810 are excellent mixers, and they have been
frequently appearing on eBay at some darn good prices. These mixers
are tailored to voice applications, not "everything" applications like
some are (Ivie I think is tailored to voice too), they have gating,
filibuster, and a logic input/output port that you can use to find out
when a particular microphone is engaged (so you can turn off all
speakers except for the one near the microphone for instance) or force
a microphone off when it is on for too long a period of time - e.g.
when the vacuum cleaner is running in the room.
Peter Monahan of iAutomate.com has a lot of experience with these
systems and subscribes to the "arm's length" rule for open air
microphones. That said it does not mean that you can't do a large
room, it just means that you need more microphones and filibuster mode
so that whenever you are talking you are not much more than an arm's
length from a microphone. For most people this becomes too expensive
to implement. That is why products that allow you to do ASR via the
telephone are even better - you pay for the microphones where it makes
sense such as the kitchen (turn on the light overhead even if you have
chicken guts on your hands), but in other rooms use the telephone.
Even if I installed a microphone at the end of the wire I ran to my
master bedroom I could rarely use it because most of the time when I
am in there my wife is asleep and would be awakened by me talking to
the house via microphone. (Her always being asleep when I come in is
a whole different problem... ;-)) I can talk into a phone with much
less chance of waking her up. I can also take the portable telephone
outside on the deck and control the house and the music being played
over the outdoor speakers without having to even think about
installing a microphone outside.
So the bottom line IMHO is that yes you can do these large rooms, but
to do it in such a way as to make it work well, it's going to cost
you.
Rick