User Guides¶
These guides will help walk you through the usage of beamforming toolkit 2.0 (BTK2.0).
In the BTK, there are two important design concepts:
- Feature pointer object that makes easy to implement a chain of signal processors, and
- Iterator of the feature pointer object for real-time processing.
In many cases, we would need to propagate the output of one audio processor to another. In BTK, a chain of speech processing can be readily implemented by cascading the feature pointer object to another. In the following sections, we will see example scripts.
Moreover, in many speech applications, a low system latency is preferred. For example, a processing delay should not be noticeable in a video conference system. Otherwise, users will find a strange lag between talker’s lip movement and voice. The low latency is typically achieved by using block-wise processing: processing a small chunk of the incoming signal in real time. For that, the BTK provides the iterator method, the next() method at C++ level and Python iterator.