Introduction

This documentation serves as an introduction to the Yobe SDK. The library requires two channels of interleaved input audio data (16 bit PCM encoded) at a sampling rate of 16kHz. The library returns one channel audio data as output with a sampling rate of 16 kHz and some additional data in some cases.

SDK Variants

The Yobe Android SDK offers the following variants:

• Near-CAFE Listening (BioListener) - Use cases when the voice of interest is closer to the microphone inputs than the background noise
• Far-CAFE Listening (BioListener) - Use cases when the voice of interest is farther from the microphone inputs than the background noise
• Near-LATTE ASR-Listening (IDListener) - Use cases when a registered speaker needs to be transcribed in an environment with noise and other un-registered speakers
• Near-LATTE AVR-Listening (IDListener) - Use cases when a registered speaker needs only their speech in the output audio and no noise nor the speech of other un-registered speakers
• GrandE (BioListener + IDListener) - Can be configured to serve all the above use cases

Depending upon your use case, you must use one of the variants mentioned above.

Note: The variants are mutually exclusive i.e. only one of them can be instantiated at a time.

Further Information

• There are functions in com.yobe.speech.SpeechUtil to assist integrating the Yobe library into your project.
• If you run into any problems, take a look at the logging capabilities in com.yobe.speech.SpeechUtil to get more information.