Blog

Current state of the art ASR system has outperformed conventional ASR system. The performance of deep neural networks in ASR has reached to professional human transcribers in clean speech environment conditions. However, it has been affected by the following challenges:

• Physical and social variances of speakers.
• Environmental and channel distortions.
• Room reverberation in far-field ASR.
• Code-switched phenomena.
• The mismatch between training and test data.

In this section; we will highlight emerging and state of the art methods used for building a robust speech recognition system from research point of view. e.g we will explain how can we address the challenges occurred in speech recognition as mentioned above.