Inverse of logamplitude?

librosa.logamplitude is great! How about also having an inverse available? E.g. something like:

def linearamplitude(log_S, ref_power=1.0)
    """Linear-scale the amplitude of a spectrogram."""
    return 10 ** (log_S / 10 + np.log10(ref_power))

I guess it wouldn’t be exact though, but still useful in some cases maybe?

[bmcfee]

Before I go hacking on this, does this sound like a reasonable set of changes?

• logamplitude renamed to power_to_db; an alias included for backwards compatibility
• new function amplitude_to_dbwhich just wraps power_to_db(X**2)
• inverse functions db_to_amplitude and db_to_power defined accordingly

[dpwe]

On Wed, Oct 19, 2016 at 10:26 AM, Carl Thomé [e-mail] wrote:

I’M CONFUSED. [image: 🐰]

So this is why logamplitude() is called with S**2 throughout the docs? To make magnitudes into power? Always been a bit confused by this actually.

Exactly.

DAn.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/librosa/librosa/issues/421#issuecomment-254829114, or mute the thread https://github.com/notifications/unsubscribe-auth/AAhs0bwicvm16_tt20NEkjp2nhYrcJNkks5q1iipgaJpZM4Ka9si .