Kung Yao - US grants

The proposed work deals with application of parallel processing to communication systems problems, mostly with particular regard to implementation with the INMOS transputer device. There are three major areas with quite distinct objectives: (1) development of a high-speed simulator for analysis of digital communication systems, (2) investigation of new architecture ( generalized lattice) for implementation of digital filters, and (3) reformulation of signal processing algorithms typically used in communication systems ( least squares estimation, Kalman filtering, sequential decoding, etc.) for a more direct and efficient implementation as a form of systolic array. The objective of area (1) is to provide an effective analysis tool for designers of complex communication systems by virtue of the high computational speeds afforded by parallel processors. Areas (2) and (3) have similar objectives and represent an investigation of a more fundamental nature. The goal is to show ways to implement sophisticated signal processing operations (such as; bandpass filtering, Viterbi detection, adaptive channel equalization) at high speed so that communication channels can be utilized more efficiently. Another benefit of this work is the development of design tools for using an exisiting device ( the transputer) for implementing common signal processing operations.

This proposal will partially support registration, travel, and hotel/per diem expenses of graduate students and recently graduated young researchers from U.S. universities and industries to attend and participate in the 1992 IEEE Workshop on VLSI Signal Processing to be held October 28-30 in Napa, CA.

This award supports the development of robust and adaptable sensor arrays and related software for observing and analyzing bird diversity and behavior. Birds have been selected as the test group for this work because of their importance in biodiversity, and because they are well-characterized with regard to the properties that will be explored (i.e. bird vocalizations). This project will make use of existing sensor arrays developed by the UCLA Center for Embedded Sensing (CENS).

The work will be performed in laboratories at UCLA and at several field sites: (1) the UC
Riverside James Reserve, near Idyllwild, CA. - a heavily instrumented field site with full-time specialists on sensor arrays. Here, methods will be initially tested, then deployed to either (2) the Hastings Reserve, near Monterey CA, where acorn woodpeckers have been marked and studied. The sensor arrays will be used to identify individual woodpeckers, locate them, and to identify patterns in their behavior associated with the approximately 10 calls they employ. Arrays will also be deployed to (3) the Montes Azules Biodiversity Reserve in Chiapas, Mexico, which has a rich diversity of bird species in a tropical rainforest. Here, the sensor arrays will be developed to identify and locate different bird species in the region.

The sensor arrays, software, and methodology developed by this project will serve a severe need to assay biodiversity in ecosystems. The tools needed to remotely sense, record and automatically analyze acoustical behavior will be enormously helpful for studies of ecology, biodiversity and behavior.

A grant has been awarded to Drs. Daniel T. Blumstein, Charles E. Taylor and Kung Yao at the University of California Los Angeles to host a workshop that will bring together current and potential users of bioacoustic sensor array technology with engineers and computer science experts to identify needs and current abilities, and to prioritize technology development and research questions.


The meeting will be held at the James Reserve, a University of California Field Station notable for the degree to which cutting edge sensor array technology has been deployed. The meeting will have a series of keynote talks and break-out groups where researchers from different fields will collaborate to develop a position paper.


The broader significance of this work is to facilitate the development of robust, easy-to-use acoustic sensor arrays with user interfaces appropriate for researchers in animal behavior and ecology, to excite promising students about the development and application of this technology, and to provide underrepresented engineering students a frontier intellectual and technical experience to encourage them to pursue advance study or to continue to the doctorate.

This award supports a collaborative effort between the PIs and investigators at the University of Arizona and Massachusetts Institute of Technology to design and develop instrumentation for wildlife biologists to inventory animals by detecting, recording, and analyzing their sounds. The system will also allow for field biologists to ask questions about the temporal and spatial dynamics of acoustic communication. Many species produce sounds, and by identifying and localizing them, we can census biodiversity and study the natural dynamics of communication. Typically, these will be vocal sounds, such as bird song, or mammal or frog calls, but in theory, the equipment and algorithms can be used for other sorts of sounds (such as the stridulations of cricket wings). To do so, we must build robust hardware and easy to use software that field biologists can and will use. While there has been the development of ?proof-of-concept? tools and algorithms for many of the components of a usable system, there is no reliable, robust, and easy-to-use system that will permit field biologists to easily census acoustic animals. To do so, a platform called VoxNet will be developed. VoxNet is an integrated software and hardware package which will be a quantum leap forward beyond existing technology in for main areas: software, near-real time event recognition, energy efficiency, and a much longer communication range.

VoxNet will be a new, highly integrated, deployable acoustic sensor node with a lower unit cost, and lower energy cost than any existing system. VoxNet will include a highly optimized system software and driver to reduce energy costs through duty cycling. In addition, new 2D/3D algorithms to localize the source of the sound will be developed. These new algorithms will run in near-real-time on a distributed network of VoxNet nodes, enabling users to detect and identify vocalizing animals while in the field. A powerful and easy-to-use software environment based on the WaveScope programming model and the XStream distributed stream processing system (technology partially developed under NSF support at MIT) will be developed. Finally, the system will be tested in the field to both census birds and identify individually alarm calling marmots.

The development of tools for field biologists to use will create new ways to census biodiversity, and ask here-to-for impossible or difficult-to-ask questions about the spatial and temporal dynamics of vocal displays. When deployed, we expect these tools to generate novel and important discoveries. The dissemination of these tools will enhance research productivity in behavior, ecology and evolutionary biology. And, these tools will create novel ways to census, conserve, and manage biodiversity. The process of developing these tools will create integrative educational opportunities for undergraduates and graduate students.

CDI-Type II. Acoustic Sensor Arrays for Understanding Bird Communication


The intent of this project is to permit humans to understand the grammar and meaning of bird songs. Recent advances in sensor arrays, computation, and computational linguistics finally make this long-sought goal achievable. The approach taken in this proposal is to: (1) collect very large amounts of bird song recordings from acoustic sensor arrays in a variety of natural settings; (2) process the data by software, some of which is recent and some of which will be developed using new advances in localizing source with beamforming, then filtering out noise, identifying events of interest, and then classifying them according to species and individual, and combining that with behavioral observations; (3) this information/knowledge will then be stored in a large database that can be shared among the collaborating research groups; and (4) it will be analyzed by computational-linguistic tools to identify the syntax of the songs, and combined with information about the context in which it occurred, then analyzed by new software methods to identify the meaning of those songs. The project begins testing inferences from those inferences and explore consequences for individual and community ecology.

The research will be transformational in several ways. First, it will contribute to a profound transformation that is already underway: the recognition of very sophisticated signaling strategies and syntactic structures in non-human species. The new tools and methods for collecting and analyzing bird song now allow a level of observation that previously would not have been possible. Scientists are now collecting truly vast amounts of data from previously inaccessible settings and subjecting data to previously undiscovered sophisticated structural analyses. It will be transformational to computational linguistics if the natural world beyond humans were shown to have languages that are radically different from our own (as seems quite likely). In addition, the project will radically expand the range of engineering with voice recognition and classification, which so far has been restricted almost exclusively to humans.

Other contributions will come from the database that will comprise huge amounts of data pertaining to bird songs and the environmental/behavioral context in which it occurs. Offering both thematic and outreach contributions, the project will bring together people from engineering, ecology, linguistics and art -- and from the US, Mexico and Japan. The educational part will bring together underserved K6-12 students with the research community and will involve them with well-established educational programs in engineering, biology and art | science. While the science portion of this project is high-payoff --- high-risk, the outreach portion will certainly be effective at furthering appreciation and learning of science.