Search

A method is described for assessing the performance of automatic tracking systems against various metrics that quantify performance in different areas. A desired performance level is defined for each metric, and the probability that a particular track will meet this performance level is empirically approximated from archived sensor data. The individual metric probabilities are combined using a weighted average to produce an overall tracker rating. An important feature of the approach is that provision is made for the case where absolute truth is unavailable, and output of the trackers under test can be compared with incomplete truth. The incomplete truth is produced by a manual inspection of the data and through editing the output of the best available tracking algorithm. An example of applying the method is presented. [ABSTRACT FROM AUTHOR]

A typical tracking algorithm takes its input from a peak detector or plot extractor. This process reduces the sensor image data to point measurements and reduces the volume of data that the tracker must process. However, useful information can be lost. This paper shows how the curvature of a peak can be a useful feature for discriminating false alarms and valid detections. The benefit obtained by using this feature is quantified through false track rate on recorded sensor data. On recorded data with difficult clutter conditions, approximately sixty percent of false tracks are rejected by exploiting peak curvature. [ABSTRACT FROM AUTHOR]

A letter to the editor is presented in response to the article "Time taken for digital dictation - examining the impact of a new hospital system" by K.A. Hanks and P.J. Clamp in the April 29, 2012 issue which discusses about the digital dictaphones with integrated barcode scanners.