dbetchkal / acousticrecord Goto Github PK

i.e., by hour, by day

these will allow better comparison to empirical data

The noise floor changes the distribution of sound pressure levels quite a bit and will be necessary to compare to empirical observations. Peaky histograms of Leq, 1s are a good example:

Shouldn't be too hard - slightly variable noise floor time series as an attribute, self.noise_floor would allow the creation of a fourth time series, self.censored_record

For LD831 data I was using noise_floor = np.random.normal(19.0, 3.0, size=rec.duration), which probably should become the default.

Received comments (on 02/10/20) related to use of the Gaussian distribution in point_source(). Potentially the Rayleigh or Beta distributions could be a better choice.