jurek
Well-known member
I am thinking mostly about Merlin from Cornell, which is probably the only practically useful software currently. However, I am interested in other possible similar software.
I would hope to have:
- Better filtering out of ambient noises. Merlin currently stops working when there is a sound of footsteps or a bicycle. It should be easy, because on the sonogram produced, low noises of footsteps or a bicycle are clearly different from high sounds of birds. But Merlin stops producing an identification.
- More species identified, or an algorithm which requires few recordings as a training set. As of today, it is said that the algorithm in Merlin needs at training set of at least 150 recordings. So many recordings are unlikely to become practically available for rarer birds. And even a common bird can have multiple very dissimilar sounds.
I would hope to have:
- Better filtering out of ambient noises. Merlin currently stops working when there is a sound of footsteps or a bicycle. It should be easy, because on the sonogram produced, low noises of footsteps or a bicycle are clearly different from high sounds of birds. But Merlin stops producing an identification.
- More species identified, or an algorithm which requires few recordings as a training set. As of today, it is said that the algorithm in Merlin needs at training set of at least 150 recordings. So many recordings are unlikely to become practically available for rarer birds. And even a common bird can have multiple very dissimilar sounds.