POLYNESIA Analysis system

The beauty of this program is that is has been built off the LabView platform, which means we can program Polynesia to record or detect virtually anything....

Polynesia, the analysis application, was created using LabVIEW 6i from National Instruments Corporation. LabVIEW is a multi-platform graphical programming system specifically designed for test, measurement, and analysis software development. Of particular importance to this application were LabVIEWs’ ease of use throughout the development cycle, the availability of portable data acquisition hardware (also from National Instruments), and the intimate relationship between LabVIEW and that hardware. The researcher has ultimate control over the form and function of the final application, and is not limited by tools that have been developed for frequencies within the human hearing range.

1. Hardware

Hardware for the analysis system consists of any portable computer with 92 Mbytes of RAM and sound card, a connector block, and a data acquisition card.

a. Connector block. Data from the recorder or microphone is first passed through a 50-pin input/output connector block, which connects field I/O signals to the PCMCIA data acquisition card (DAQ). The National Instruments CB-50LP connector block is housed in a 140 mm x 80 mm waterproof box that weighs .35 kg. A 50-pin cable assembly, the National Instruments PR5050M, attaches the connector block to the PCMCIA DAQ card.

b. Data acquisition device and DAQ software. Data acquisition can be facilitated using I/O data acquisition equipment adapted for use with portable computers such as PCMCIA cards, parallel port, or USB-Firewire devices developed by National Instruments. Some of these have sampling rates of 1.25 MS/s. The value of this flexibility is that the researcher has complete control over the sampling speed, which is entirely dependent on the type of data acquisition hardware used, as noted in the LabVIEW Measurements Manual (2000).

For the tiger study, the primary data acquisition hardware was the DAQCard-1200, a multifunction PCMCIA card that has eight analog input channels, two analog output channels, and one digital port. The DAQCard-1200 performs up to 100 kS/s, with 12-bit resolution, which allows the researcher to analyze sounds up to 50 kHz (LabVIEW Measurements Manual, 2000). There are cards now available that sample at 500 kS/s.

The hardware and I/O channels are managed using the Measurement and Automation Explorer (MAX) a high level program used to configure National Instruments hardware and software. This configuration-based approach to I/O management allows software development to proceed - based on any desirable measurement unit, while affording the researcher the freedom to investigate new and different sensors that may have widely contrasting gains, calibration factors, and hardware requirements. "Virtual Channels" are named and configured within the Data Neighborhood folder in MAX. Embedded in that channel configuration is all the information necessary for LabVIEW to make a measurement and supply data in the appropriate measurement unit to the application. This information includes the physical address of the signal input, the sensor type, and all scaling and calibration data for the sensor. Scaling and calibration data can be in the form of a linear map, a polynomial or, as specialty sensors often require, a vendor supplied look-up table. The result is that the sensors and the data acquisition hardware can be changed at will by the researcher with no need to touch the source code of the application (LabVIEW Measurements Manual, 2000).

The application is also capable of I/O through a standard PC sound card, though such I/O is of limited usefulness outside the range of human hearing. It should also be noted that this device is controlled through the Windows API and, therefore, has none of the benefits or conveniences offered by MAX.

 

2. Software

The User Interface is divided into three general areas: The display of real-time data; a display used to capture and refine coarse snapshots of the real-time data; and a section used to perform analysis, and edit the snapshot data for playback and storage.

a. Real-time data. The top third of the screen contains the menu bar and two real-time data displays, a time domain chart, and a frequency domain graph. The primary menu functions select which input or output to use and control the sample rate of the input. The sampling rate is typically used from 4 kHz to 44.1 kHz, but can be set virtually anywhere up to 50 kHz, where 100 kHz is the upper limit of the DAQCard-1200. The chart displays scrolling amplitude data versus time, and the graph is a continuously updated Auto Power Spectrum of the amplitude data. The power spectrum is computed as F*[V(t)F [V(t)]/ N2, where F[V(t)] is the Fourier transform of the time-varying signal V(t), where N denotes the number of points in the signal array, and * denotes complex conjugate. The power spectrum is then converted into a single-sided power spectrum result (Chugani, 1998

b. Snapshot function. The middle third of the screen contains a control used to take a "snapshot" of the live data and a time domain graph to display that data. This graph has two vertical cursors that are used to mark the beginning and end of significant events to analyze or copy to the manipulation area of the screen. In practice, the researcher watches the real-time time domain chart for an event, then clicks the Snapshot control while the event is scrolling through the display. This action transfers the contents of the time domain chart buffer into the Snapshot graph. When a tiger roared or made a growl or other vocalization, snapshots of the signal were taken. The graph cursors were then used to further isolate the event.

c. Analysis and editing. The bottom third of the screen is implemented in a tabbed control to conserve screen real estate and to provide logical grouping of the analysis and manipulation functions. There are three different types of functions in this section: Fast Fourier transform; spectrographic function; and signal editing functions including cut, paste, high and low band pass filtering, saving and signal output. Both the FFT and the spectrographs can analyze at rates between 0 and 50 kS/s. Although the default is Hamming, the user can choose between Blackman-Harris, Exact Blackman, Blackman, 4-term Blackman-Harris, 7-term Blackman-Harris, and Hanning windowing (Cerna and Harvey, 2000).

The first tab, "1D," performs an FFT Power Spectrum on the data between the Snapshot graph cursors, as shown in Fig. 1. The user can set the maximum and minimum frequencies on the graph to zoom in on a particular band of frequencies after the analysis has been performed (Chugani, 1998The second tab, "2D," performs a spectrograph, short-time Fourier transform, or STFT (Chugani, 1998) on the data between the Snapshot graph cursors. The display of this spectrograph is an intensity graph where time is in X, frequency is in Y, and magnitude is in Z, and is represented visually as color variations through a predefined color "palette" as shown in Fig. 2. The STFT is rather processor intensive, so live data acquisition and displays are suspended for a few seconds during this operation.

 

The third and final tab, "Misc", (figure 3) contains another graph, the "Workspace", with two cursors and a number of buttons used to manipulate the data in the Workspace, (LabVIEW custom controls, 2000). The researcher can copy data from the Snapshot graph, copy and paste data (signal editing) within the Workspace, delete portions of the workspace data, or clear the graph entirely, and perform cross correlations with cepstrum. A filter can be applied to the Workspace data. This filter is configured in a pop-up window and can be highpass, lowpass, bandpass, or bandstop. An FFT can be performed on the Workspace data to observe the results of filtering. This FFT is displayed in the graph on the 1D tab (LabVIEW analysis concepts, 2000). Crosscorrelation with cepstrum can alsobe performed. After signal editing, the Workspace data can be saved to a standard WAV file and WAV files can be recalled into the workspace. Finally, the data can be played back through the PC sound card or routed out through one of the analog outputs of the data acquisition device to speakers. When using the computer sound card output, the researchers can double or triple and/or expand the playback speed to make the lower/higher frequencies audible. When using the data acquisition device output, the sound can be enhanced to virtually any playback speed, so even a 7 Hz signal can become audible to the human ear.