Hi guys.. my project is to detect sounds from babies inside vehicles, once a voice is detected, trigger an alarm.. so i was wondering, would it be possible to use an "electret" microphone and set it in the arduino to detect a certain range of human frequencies, since if the vehicle is in a busy street, it could just take any sound(car horns,construction site noises) and trigger the alarm. Im pretty sure ive heard that about range of frequencies in human voices, and male,female and even babies are all different. If it is possible, i would like to buy the microphone and get started on it.. Or is there any other better way to do this, and if so could anyone point me to an example, since ive not found any site that is reliable. TQ
You will have to build an active band-pass filter with/and and amplifier. (not to hard)
e.g. http://www.electronics-tutorials.ws/filter/filter_7.html
I think the biggest problem will be to select the frequency range for best detecting results.
Most of the Electret come with a built in 100 amp, so the real problem would be to sort out baby from all the other loud noise. But why, you can't leave the kid alone in the car?
Mark
Sorry but there is no way that is going to work. The range of frequencies is not exclusive to the human voice all other stuff uses it as well.
?
Grumpy_Mike:
Sorry but there is no way that is going to work. The range of frequencies is not exclusive to the human voice all other stuff uses it as well.
holmes4:
Most of the Electret come with a built in 100 amp, so the real problem would be to sort out baby from all the other loud noise. But why, you can't leave the kid alone in the car?Mark
im suppose to design a safety feature for cars, where parents shouldnt leave their babies in there...the microphone are one of the features aside from PIR..so what your saying is there is no good way in detecting the frequency of human voices? i have to use a microphone cause it is a requirement set by my lecturers, what do u think i should do, to differentiate human voice ( babies crying ) from other enviromental factors
knut_ny:
You will have to build an active band-pass filter with/and and amplifier. (not to hard)
e.g. http://www.electronics-tutorials.ws/filter/filter_7.html
I think the biggest problem will be to select the frequency range for best detecting results.
hey there man, nice to see you again, yeah, i was thinking that it would be an analog signal, so i figured i could just insert certain range of numbers, and trigger the alarm, but now i guess there's no possible way to detect human frequency?
Analog telephone lines used BP-filter approx 300-3500 HZ. This is good enouch to distinguish most voices from another.
see: Voice frequency - Wikipedia
A childs normal voice would be (talking) in tre range 200..400Hz When crying I guess higher frequencies.
For a test: Start out with the mic-signal through amp -> filter -> rectifier.
Let the rectified signal charge a capacitor, where you tap out to analog in
(this C must have a parallell (large value)R- "to leak charge to ground")
knut_ny:
Analog telephone lines used BP-filter approx 300-3500 HZ. This is good enouch to distinguish most voices from another.
see: Voice frequency - Wikipedia
A childs normal voice would be (talking) in tre range 200..400Hz When crying I guess higher frequencies.
For a test: Start out with the mic-signal through amp -> filter -> rectifier.
Let the rectified signal charge a capacitor, where you tap out to analog in
(this C must have a parallell (large value)R- "to leak charge to ground")
thanks for the advice man, haha sorry to be a burden but is the electret microphone good enough? cause im trying to buy one after this..can it do what u said?
and when u said "(this C must have a parallell (large value)R- "to leak charge to ground")" , what do u mean?sorry again..still new
..i'll have one of my students do a "test-build".. then report results.. (within few days)
so what your saying is there is no good way in detecting the frequency of human voices?
No a microphone is a perfectly acceptable way of detecting the frequency of human voices. However that is not saying that if you detect a frequency in that range that it actually is a human voice, it could be absolutely anything from a car horn engine noise the air conditioning, they all have frequency components in the same range as a human voice.
Think about this:-
All fish swim in water, all things that swim in water are not fish.
it is a requirement set by my lecturers,
So this is a project you are supposed to do, if you are asking questions this fundamental then I guess you are not expected to do the heavy signal processing that might just give you a glimmer of a chance at distinguishing a human voice from other noise. Doing that is not a simple matter of a band pass filter.
I've been thinking about this. It may not be possible the detect the babies cries, but we could rule out a lot of the other things.
-
Inside a parked car its normally quite, very quite compared with a crying baby.
-
the pitch of the cry
-
The duration babies cry and cry and cry and ...........
Combine the 3 and you rule out a lot of things and it the closed enviroment of the car what else could it be?
Mark
I think we can be comfortable with this as a purely academic exercise without repercussions. 
A parked car will be quiet only if the windows are closed - which is of course, all the more dangerous. So the absence of ambient noise is an added criterion and increases effective discrimination.
Sleeping babies do not cry - and may not wake with heat exhaustion.
A suitable characteristic for recognition would be the pitch and envelope - progressive attack over about half a second and slightly longer decay, duration approximately 2 seconds in general.
Repeated.
An electret microphone is pretty much standard nowadays - used on all mobile phones, laptops and similar devices.
So assuming we're at a busy street there's no possible way to detect their cries? think it would be much more complicated than what i can manage as far as my knowledge goes..
Paul__B:
A suitable characteristic for recognition would be the pitch and envelope - progressive attack over about half a second and slightly longer decay, duration approximately 2 seconds in general.Repeated.
An electret microphone is pretty much standard nowadays - used on all mobile phones, laptops and similar
When you say using pitch , can u explain what you mean by that? The pitch given of from the crying? How does tgat work? Can u please explain to me a bit and i will read more on it
Pitch is another word for frequency. You have a lot to learn.
Grumpy_Mike:
Pitch is another word for frequency.
Except perhaps that it is used in respect of musical(?) tones containing overtones, which may complicate the analysis (or may even be critical to the analysis).
Paul__B:
Grumpy_Mike:
Pitch is another word for frequency.Except perhaps that it is used in respect of musical(?) tones containing overtones, which may complicate the analysis (or may even be critical to the analysis).
No, pitch is frequency. When talking about music overtones do not contribuites to the pitch but affect the tambura of a note. That is how it sounds, like a trumpet and a violin can play at the same pitch but they don't sound the same because they have differant tambura.
..We had no child to beat up, so they used "audacity" to generate test frequencies.
The design needs a rail to rail opamp to work. Pot-meter is linear, but any pot 10k+ will do.
Gain must be adjusted with the feedback resistor.
With sustained high freq. noise in, a voltage will bild on 10uF (leeking throud 1M) The comparator gives logical HIGH when this voltage exceeds preset V
Well that is just an amplifier and an envelope generator. What makes you think it will distinguish a child's cry from any other noise?
didnt say it will. It detects 1kHz+ above selectable SPL during a time period (few seconds).
We agree that simple electronics cant replace moms ear..