Archived project

Gisting Conversational Speech

Goal: The project was to extract pilot-flight controller dialogs from air traffic control recordings. I worked on the dialog identification code which used unsupervised agglomerative clustering based on cepstral features to identify the separate controller and pilot conversations. The 1st step was to separate the pilots from the controllers which was easy due to pilots have more background noise. Then we did cuts lower in the dendrogram to pull out the separate pilots in the pilot side of the tree and controllers from the controller side of the tree. I wrote an algorithm to then intersect the controller clusters with the pilot cluster to form dialogs. I also wrote the user interface to the overall system on a Symbolics Lisp Machine and used Ken Anderson's outstanding Grapher package to generate the many graphs and graphical drill downs in the interface.

Updates
0 new
2
Recommendations
0 new
0
Followers
0 new
0
Reads
0 new
9

Project log

Albert Boulanger
added an update
Here is a presentation on the cluster intersection algorithm I developed to form dialogs.
 
Albert Boulanger
added a research item
A novel system for extracting information from stereotyped voice traffic is described. Off-the-air recordings of commercial air traffic control communications are interpreted in order to identify the flights present and determine the scenario (e.g., takeoff, landing) that they are following. The system combines algorithms from signal segmentation, speaker segregation, speech recognition, natural language parsing, and topic classification into a single system. Initial evaluation of the algorithm on data recorded at Dallas-Fort Worth airport yields performance of 68% detection of flights with 98% precision at an operating point where 76% of the flight identifications are correctly recognized. In tower recording containing both takeoff and landing scenarios, flights are correctly classified as takeoff or landing 94% of the time
Albert Boulanger
added a project goal
The project was to extract pilot-flight controller dialogs from air traffic control recordings. I worked on the dialog identification code which used unsupervised agglomerative clustering based on cepstral features to identify the separate controller and pilot conversations. The 1st step was to separate the pilots from the controllers which was easy due to pilots have more background noise. Then we did cuts lower in the dendrogram to pull out the separate pilots in the pilot side of the tree and controllers from the controller side of the tree. I wrote an algorithm to then intersect the controller clusters with the pilot cluster to form dialogs. I also wrote the user interface to the overall system on a Symbolics Lisp Machine and used Ken Anderson's outstanding Grapher package to generate the many graphs and graphical drill downs in the interface.