Using Mic from 10 feet away for voice recognition

Question

I'm trying to use a mic from 10 feet away.

Background: for my senior project I would like to talk to a computer to run scripts, but I want to talk to the mic throughout the house. So I need to be able to speak from about 10 feet away to a mic in the ceiling, and have it go to the computer, where the computer will execute the scripts based on the voice commands.

Due to my inexperience with op amps and audio signal design, what would be a good way of amplifying the signal, and what would be a good way of filtering the signal?

The Quality is important, because the voice recognition will work better with better quality mics, and I might be talking over the phone with the mic. Budget is also important. because i have to build 4 to 5 of them.

Any Suggestions also for how to approach the problem are also welcome. Thanks

Answer 1

Since your question is very general, I can only guess as to what guidance you actually need. Here's a start.

Due to my inexperience with op amps and audio signal design, what would be a good way of amplifying the signal, and what would be a good way of filtering the signal?

Amplification and filtering often go by the single term signal "conditioning".

Given that the human voice has a range of about 300Hz to 3.4kHz (according to Wikipedia), it would seem that a bandpass filter would be sufficient to reject frequencies outside this range. An active bandpass filter, one with active electronics (usually op-amps), allows you to amplify your signal as well (since op-amps are just that, amplifiers). This could take some of the weight off a separate amplifier circuit, which you may need to boost your signal to usable levels.

Here's a good link to active bandpass filter topology and analysis. This link makes good arguments for both cascaded filter design (first section) and multiple feedback design (second section).

Cascaded design uses three stages which combine to create an active bandpass filter: a highpass filter, an amplifier, and a lowpass filter. This approach gives you a very flat passband, which makes it good for wide ranges like our 300Hz to 3.4kHz. If you implemented this method, I would perform the lowpass filtering first, because the most prominent noise in your system will likely have high frequency content, and you don't want to amplify that before filtering it.

The multiple feedback design also has its pros. For example the issue with doing lowpass filtering first (described above) is negligible. However, it is best used for narrow passbands, which 300Hz to 3.4kHz is not. I'd recommend the cascaded approach above.

Answer 2

The problem with a microphone 10 feet away for voice recognition isn't just volume or amplitude, but also multi-path interference (or room reverb). Sound reflecting off of the walls, floors, furniture, etc. is much weaker than your voice when the mic is up close. Not so when the mic is far away. This multi-path interference can change and even cancel out portions of the spectrum of your voice, and in a manner that changes anytime you move or face a different direction, which makes voice recognition more difficult.

Solutions might include using multiple microphones and DSP processing to determine and partially cancel any multi-path distortion, or perhaps robotically aimed directional microphones that can follow your position in the room.

Answer 3

if the mic is 'far away' from your recording device, you need a low impedance signal,

for short: build a amp en place it as near as possible by the mic. a mic to line amp / mic preamp circuit is suitable for that.

the amplified signal goes to in your case your computer. also like Russel McMahon says: "directional microhones" & "shotgun microphones"; you can place more directional mic's and sum them together, or use a shotgun mic.