AAC | ADPCM | Convolution | DAFx | FFT | IIR | Mixer | MP3 | MPEG | MPEG-4

Hello,

I am currently using a digital 1000 point lowpass filter to extract 
the envelope from an audio signal of which I have removed the 
negative components (by using the absolute value of the signal). I am 
using this resulting envelope to detect whether notes have occurred 
or not.

I would like to investigate better filter's responses to extract this 
envelope from an audio signal. The next filter I will test is the 
exponential moving average IIR filter.

Can anyone suggest a better filter to extract the envelope from an 
audio sampled 44.1kbps signal? 

I would need the filter to be responsive to changes quickly and as 
closely represent the actual audio as possible.

I understand there is tricky modified averaging filters used in the 
stock market/financial fields but unfortunately I have no experience 
with any of these methods.

Any suggestions would be greatly appreciated.

Regards,
Brett Carruthers

Systems Engineer
iiiTrisys

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )

Hello,
>
>I am currently using a digital 1000 point lowpass filter to extract 
>the envelope from an audio signal of which I have removed the 
>negative components (by using the absolute value of the signal). I am 
>using this resulting envelope to detect whether notes have occurred 
>or not.
>
>I would like to investigate better filter's responses to extract this 
>envelope from an audio signal. The next filter I will test is the 
>exponential moving average IIR filter.
>
>Can anyone suggest a better filter to extract the envelope from an 
>audio sampled 44.1kbps signal? 
>
>I would need the filter to be responsive to changes quickly and as 
>closely represent the actual audio as possible.
>
>I understand there is tricky modified averaging filters used in the 
>stock market/financial fields but unfortunately I have no experience 
>with any of these methods.
>
>Any suggestions would be greatly appreciated.
>
>Regards,
>Brett Carruthers
>
>Systems Engineer
>iiiTrisys
>

Hi
Your first method seems to be quite cpu expensive. A simple moving average
filter should do it, I think (averaging the squared or absolute values of course).
I think of something like this:
y[k] = y[k-1] +   1/n*(x[k]^2 - x[k-n]^2)

gr.
Anton

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )

Brett (and also Anton)-

> I am currently using a digital 1000 point lowpass filter to extract
> the envelope from an audio signal of which I have removed the
> negative components (by using the absolute value of the signal). I am
> using this resulting envelope to detect whether notes have occurred
> or not.
> 
> I would like to investigate better filter's responses to extract this
> envelope from an audio signal. The next filter I will test is the
> exponential moving average IIR filter.
> 
> Can anyone suggest a better filter to extract the envelope from an
> audio sampled 44.1kbps signal?

I might suggest researching use of Hilbert transform to find the envelope of a time
domain signal.  Here is one page that gives a simple explanation:

  http://www.numerix-dsp.com/envelope.html

The idea is that 90-deg phase shift gives you a better starting point than just
rectification before you apply sum and hold.

-Jeff

> I would need the filter to be responsive to changes quickly and as
> closely represent the actual audio as possible.
> 
> I understand there is tricky modified averaging filters used in the
> stock market/financial fields but unfortunately I have no experience
> with any of these methods.
> 
> Any suggestions would be greatly appreciated.
> 
> Regards,
> Brett Carruthers
> 
> Systems Engineer
> iiiTrisys

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )