Verification method for Chinese aviation radiotelephony readbacks based on LSTM‐RNN
Authors:
Open Access
- 1 March 2017
- journal article
- information and-communications
- Published by Institution of Engineering and Technology (IET) in Electronics Letters
- Vol. 53 (6) , 401-403
Abstract
Reading back of the instructions acquired by pilots through radiotelephony communication from air traffic controllers plays a very important role for civil aviation safety. Whereas the mistakes of readbacks are difficult to find out when the controller or the pilot is under great pressure, fatigue, tension etc. To solve this problem, the authors propose a novel semantic consistency verification method based on recurrent neural network with long short‐term memory structure (LSTM‐RNN) for Chinese radiotelephony readbacks. The actual Chinese civil aviation radiotelephony recordings are converted to textual format, and the semantic similarity is studied to verify whether the semantics is the same between the controller instructions and the pilot readbacks. The word‐based feature is extracted by one‐hot vector, and LSTM‐RNN is employed to build up a deep network architecture for producing high‐level sentence semantic abstraction of the initial input instructions and readbacks pairs. Cosine similarity is used to quantify the semantic similarity, and different classification methods are adopted to verify consistency in semantics. The experimental results show that the method is effective and provides a new scheme for the intelligent checking of aviation radiotelephony readbacks.Keywords
Funding Information
- National Natural Science Foundation of China (61379102)
This publication has 5 references indexed in Scilit:
- Deep Sentence Embedding Using Long Short-Term Memory Networks: Analysis and Application to Information RetrievalIEEE/ACM Transactions on Audio, Speech, and Language Processing, 2016
- Cosine distance features for improved speaker verificationElectronics Letters, 2015
- Long-Term Wind Speed and Power Forecasting Using Local Recurrent Neural Network ModelsIEEE Transactions on Energy Conversion, 2006
- An analysis of fishing vessel accidents in fishing areas off the northeastern United StatesSafety Science, 2005
- Learning to Forget: Continual Prediction with LSTMNeural Computation, 2000