What's the meaning of bottleneck features?

[dong-8080]

"The bottleneck feature extractor used in this paper is a speaker-independent LSTM model for ASR". The above quote appears in your article.I don't quite understand the meaning of bottleneck features extractor and can't find the corresponding part in the code. Could you explain to me how to implement this model?
Thanks and regards.

[hellolzc]

The model used to extract the bottleneck features is derived from an acoustic model for Automatic Speech Recognition. The acoustic model outputs the posterior probabilities and the BN features are the output of the hidden layer in front of the output layer. You can use other similar open-source ASR models to get this feature.