Software takes new dimension as scientists develop a bird song decoder that automatically identifies the call of a vast variety of birds. This fantastic software brings the cracking of the dawn chorus a step closer and recordings of individual birds and dawn choruses were used to identify the characteristics of each tweet.
According to the reports obtained, the person concerned who developed these mighty software provided a set of recordings, each of them having a list of the bird species present and he was able to train a computer software with these labelled examples. This is known as “feature learning”, a technique that automatically analyses the respective features.
Songs of many birds were analysed, including samples from British Library Sound Archive. The correct and perfect results were compiled and it was more than the biggest dataset ever gathered.
Dr Dan Stowell is the man behind such a development and he said in his words that birdsong is most complex of all sounds and most of them sound very close enough. He believes that the existing software is not desirable as it is not specific enough. With this new technique, bird sound can be classified to a better and a very large scale.
Presently Dr Stowell is working on developing technology that can surpass the present one and can domuch more than only identifying the bird. He looks forward to develop a technique and try it out efficiently to decode the actual relationships between the birds. With analysing the call of a species of goldfinch he is even closer to get to know which birds are in pairs.
