1.1 Features

Continuous Speech Processing (CSP) is software that supports development of host-based automatic speech recognition (ASR) applications on specific Dialogic boards. It provides many features such as high-performance echo cancellation, voice energy detection, barge-in, voice event signaling, pre-speech buffering, full-duplex operation and more.

CSP consists of a library of functions, device drivers, firmware, sample demonstration programs and technical documentation to help you create leading-edge ASR applications. It is a significant enhancement to existing Dialogic echo cancellation resource (ECR) and barge-in technology.

Key features of CSP include:

• Full-duplex operation which means the capability of simultaneously sending and receiving (playing and recording) voice data on a single CSP channel.
• Echo canceller that eliminates up to 16 milliseconds of echo in the incoming signal.
• Voice activity detector (VAD) that determines when significant audio energy is detected on the channel and enables data to be sent only when speech is present, thereby reducing CPU loading.
• Voice event signaling capability which means that when the VAD detects significant energy in the incoming signal, the CSP firmware can optionally send a message to the host application.
• Barge-in capability which allows a party to speak or enter keypad digits without waiting for the end of a prompt.
• Pre-speech buffering that can store up to 250 milliseconds of incoming speech in a circular buffer, reducing the problem of clipped speech and increasing recognition accuracy. This buffered data is passed on to the host application along with subsequent speech signals.
• Ability to modify certain VAD parameters on the fly. You can modify certain VAD parameters, such as the speech threshold, while streaming or recording is in progress.
• Ability to generate both TDX_BARGEIN and TDX_PLAY events when a prompt is interrupted, rather than just TDX_BARGEIN event.
• Ability to rearm or re-enable the VAD. This is useful when non-speech such as a cough is determined.
• Ability to send an external reference signal (echo-reference signal) from another device across the TDM bus to the CSP voice channel. Using this feature allows you to share the echo canceller and VAD resource on one CSP voice channel with other devices.
• Demonstration program that illustrates the key features of CSP.

The following features are available on DM3 boards only:

• Echo canceller convergence event notification - ability to send an event to the host application when the echo canceller has converged; that is, the echo component has been significantly reduced.
• More powerful voice activity detector (VAD) - the VAD performs sophisticated calculations using a combination of energy and zero-crossing mode (where the energy level goes to zero for a time period) to accurately determine the start of speech.

For a description of technical terms, see the Glossary.

Click here to contact Dialogic Customer Engineering