Recognising musical notes

[nyk]

Hi all!!!

I wish to detect musical notes from audio sound.
For this I plan to have an analog part to first take the analog audio wave from microphone or line-in and then pre-process it.
In this pre-processing part as I have thought firstly an amplifier to amplify the low signal levels, then a filter to get the frequency band of interest
which should be from 0 Hz to say 10 KHz, then an ADC to convert the filtered output to digital samples.
After this pre-processing , the samples are processed by a DSP.

Can you suggest if the above approach in the analog part is good enough( to remove all noise) and is there a better approach so as to make the
processing in the DSP easier, or a fully analog processing based approach.
The aim is to get accurate representation of music waves from surrounding in form of digital samples.
The input is bound to acquire noise from surroundings(microphone captured audio) and from other means.

Help analog and audio guys!!!

 

[ninju]

try MFCC.

Mel-frequency cepstrum - Wikipedia, the free encyclopedia

 

[nyk]

Will try with that but I needed advice regarding the pre-processing of audio sound waves using analog methods.

How to approach the pre-analog part?? It should be robust enough to reject all unnecessary components and give only desired music information as digital samples.
Help!!!!

 

[BradtheRad]

You may be aware that you can play audio directly into a computer.

Some computers have a built-in microphone. However it's possible to pick up computer noise at the same time. Better to use a plug-in mic, webcam, etc.

There are a variety of sound recording programs that do this. Some freeware, some priced. (I use a free program called Audacity.)

Or extract audio from files which were originally recorded using a camera, cell phone, etc.

From that point your audio is in digital form.

You can process it digitally through a hand-designed filter curve.

Amplify the signal.
Extract noise.
Emphasize or deemphasize a frequency band.
Examine the frequency spectrum from moment to moment.
Observe each pitch played by a voice or instrument.
Observe patterns of overtones and harmonics.

============================

When recording is underway, you must keep the signal in a certain volume range. Too loud and the digital recording becomes staticky. Too soft and noise predominates.

It's important to do test recordings and adjustments, until you're sure the input signal is at a suitable level.

 

[milind.a.kulkarni]

I will like to share here with the best of my knowledge....when you are dealing with music....remember....Music is a singal in which perodic frequency component are have more predominent information where as in case of speech it is aperodic frequency component....generally MFCC ( mel-frequency cepstrum) is used in voice recongnization ( Speech related problems) ....where as DTFT ( Disccrete time Fourier transform ) is tool used for decompostion of musical signal.....

Good Luck

 

[wa1kij]

It wouldn't work with chords. The frequency recognition would be confused by the beat (heterodyne) frequencies.

 

[nyk]

I need to make a standalone module. So first capturing sound from mic is necessary. Also after capturing I need to convert it "myself" not using computer but through ICs or components. So this part of the system has to be analog. Is it clear??
I need help in designing this part. Analog?Music guys any idea??

 

[BradtheRad]

What you describe can be done with standalone components.

1. Mic

2. volume control (potentiometer)

3. pre-amp (probably op-amp, automatic gain control recommended)

4. consider adding a meter, or led bargraph.

5. low pass filter (to 10 kHz, probably op-amp)

6. ADC (operating fast enough to output over 20K samples per second)

7. Device which can accept bytes of data, in sync with ADC, and

8. store the data in megabytes of ram or other storage device

========================

Uninitiated users must be instructed to:

* hold the mic about 12 to 24 inches from the mouth ('pop' filter screen is recommended)

* avoid making hand-rubbing noises against the mic

* avoid making clothes rustling noises against the mic, etc.

* adjust the volume high enough

* also keep volume low enough so it never exceeds limit at the ADC.

Allowing audio to exceed ADC limits makes the result unlistenable, and unusable for DSP. Hence the recommendation to use AGC in the pre-amp.