Approach

During its initial stages, our project has roughly four principal sub-projects: First, to develop the ability to recognize individual animals from their vocalizations; second, to develop the ability to localize individuals from their vocalizations; third, to apply these abilities to scientific problems in field studies; and fourth, to engage K-12 students with the projects to further their science education.

Systems/Experiments

(1) Develop the ability to recognize individual animals from their vocalizations.

We have taken several approaches toward identifying birds or other animals that are vocalizing.  The simplest is to calculate the power spectrum, whereby the amount of energy at each wavelength is calculated and used to form a vector, typical to that individual or species. Next in power is to prepare a sonogram of the vocalization, then look at particular features of those sonograms that might be particular to the species or individual. The most sophisticated is to adapt methods for human voice recognition, especially Hidden Markov Methods (HMM), to create a Markov Transition Matrix appropriate to the vocalizations of each individual or species.

Our primary foci for these activities are acorn woodpeckers for individual identification, Mexican antbirds for species identification and, for reasons described momentarily, yellow-bellied marmots. During the past months we have obtained field recordings of these species and have analyzed  them for sonogram features.  HMM methods require many copies of each individual's vocalizations in order to construct the transition matrices.  With these methods we are currently able to identify individual acorn woodpeckers who are most vocal in our study area, the common species of antbirds in our Mexican study area, and approximately 20 marmots for which we have adequate samples.

(2) Develop the ability to localize individuals from their vocalizations.

We have adapted methods from defense applications for localizing tanks by seismic information to localizing birds by acoustic information. (The speed of propagation of seismic information from vehicles is approximately the same as the speed of sound, so the methods carry over reasonably well.) Some additional work toward creating approximately maximum likelihood methods for location was, however, necessary.

In our first study we investigate design, analysis, and testing of acoustic arrays for localizing acorn woodpeckers using their vocalizations.  Recorded "Waka" vocalizations were played through a speaker located with the range of an acoustic array.  Each acoustic array consists of four microphones arranged in a square. All four audio channels within the same acoustic array are finely synchronized within a few micro seconds. We apply the approximate maximum likelihood (AML) method  to synchronized audio channels of each acoustic array for estimating the direction-of-arrival (DOA) of the woodpecker vocalizations. The woodpecker location is estimated  by applying least square (LS) methods to DOA bearing crossings of multiple acoustic arrays. We have explored the critical relation between microphone spacing of acoustic arrays and robustness of beamforming of woodpecker vocalizations. Woodpecker localization experiments using robust array element spacing in different types of environments were conducted and compared.

The theory for these localization experiments has been developed only for two dimensions. We are starting the extension to 3 dimensions. In the meantime, so that we may begin field work, we are starting to work with warning calls of yellow-bellied marmots, working with Dan Blumstein at the Rocky Mountain Biological Laboratory near Crested Butte, Colorado. These live in open areas and their vocalizations have been much studied. Because they live in effectively two dimensions while on the surface, we anticipate that a significant proportion of our time during the coming year will be to develop ability to localize them and to explore several of the parameters that might effect our ability to do so.

(3) Apply these abilities to scientific problems in field studies.

Acorn woodpeckers: During the past months we have recorded vocalizations from a number of identified individual birds. We have done feature analysis of their songs and for several birds constructed transition matrices for HMM analyses. This will continue.

Mexican antbirds: Unlike the acorn woodpeckers, which are individually banded, the antibirds must be identified to their territories before recordings can be assigned. Also, unlike the UCB Hastings reserved, where the woodpeckers have been studied for years, the infrastructure and knowledge base at Montes Azules is less extensive. We first obtained the necessary permits and then identified the species of antbirds present there.  We then mapped our the  territories for 4-12 individuals from each of the species resident in the Chajul field station. During the breeding period we obtained recordings from those individuals, and will analyze them during the coming months. Field trials in Mexico indicate that we have a substantial, but still not adequate, ability to localize singing birds in the forest. Just what are the limiting factors are still not clear; determining that is a major focus of our current efforts.

Yellow-bellied marmots: As described above, we plan to devote some of our efforts to acoustic arrays for identifying and localizing marmots. There are a number of open questions that we believe can be answered with the arrays, especially those involving the reliability and attention given to individual callers, as a function of where they are. There is quite a bit known about their vocalizations.  We have obtained recordings of their calls; these have been analyzed by feature analysis, and we are now creating transition matrices for the HMM identification.

(4) Engage K-12 students with the projects to further their science education.

The initial education focus has been to construct a curriculum  that will integrate the potential of the acoustic sensory arrays with the topics mandated by the state of California. Such a curriculum was drafted and several experiments are being tried out.  The most ambitious of these has been the construction of a targetable web camera at the UCR James Reserve. Currently it is just directed at the nest, compatible with our original intentions. The intention is that when individual calls can be localized, then the web camera will be directed toward the caller.  Features of this will be explored during the late summer/early autumn.

All acorn woodpeckers at the UCB Hastings reserve have been marked and have been the subject of study for many years.  We have made recordings of the woodpeckers who call the most and have developed the ability to identify them, as described above.  One problem has been that the internet bandwidth to this reserve has been narrow and slow.  We are in the process of obtaining what we believe will provide us to make real-time observations and stream them to the James Reserve and/or UCLA, which will be essential for developing the full K-12 education potential that we foresee. Among other things, this will enable us to install and monitor web cameras to observe nestling behavior.

Associated with our studies in the Mexican antbirds we have begun to make recordings of the other common species in this area.  One of our co-PIs (Martin Cody) has been associated with the use of such recordings to teach local children about the main species of birds in their area of Nicaragua. The goal is to raise awareness and appreciation of biodiversity in fragile tropical rainforested areas.  A CD they prepared is now widely used in the schools there.  The administrators in the Montes Azules Rainforest Preserve are enthusiastic about preparing such a CD for the area near our Mexican Field station in Chiapas, as well. We have begun to make such recordings and hope to eventually hope to develop this into an education/outreach tool for the Montes Azules area.

Accomplishments

We are currently able to identify individual acorn woodpeckers who are most vocal in our study area, the common species of antbirds in our Mexican study area, and approximately 20 marmots for which we have adequate samples

We have explored the critical relation between microphone spacing of acoustic arrays and robustness of beam-forming of woodpecker vocalizations. Woodpecker localization experiments using robust array element spacing in different types of environments were conducted and compared.  The theory for these localization experiments has been developed only for two dimensions. We are starting the extension to 3 dimensions.