22.2433GPPRelease 17Speech recognition framework for automated voice servicesStage 1TS
Definitions and abbreviations used in the present document are listed in TR 21.905 . For the purposes of this document the following definitions and abbreviations apply:
Automated Voice Services: Voice applications that provide a voice interface driven by a voice dialog manager to drive the conversation with the user in order to complete a transaction and possibly execute requested actions. It relies on speech recognition engines to map user voice input into textual or semantic inputs to the dialog manager and mechanisms to generate voice or recorded audio prompts (text-to-speech synthesis, audio playback,). It is possible that it relies on additional speech processing (e.g. speaker verification). Typically telephony-based automated voice services also provide call processing and DTMF recognition capabilities. Examples of traditional automated voice services are traditional IVR (Interactive Voice Response Systems) and VoiceXML Browsers.
Barge-in event: Event that takes place when the user starts to speak while audio output is generated.
Conventional Codec: The module in UE that encodes the speech input waveform , similar to the encoder in a vocoder e.g. EFR, AMR.
Downlink exchanges: Exchanges from servers and networks to the terminal.
Dialog manager: A technology to drive a dialog between user and automated voice services. For example a VoiceXML voice browser is essentially a dialog manager programmed by VoiceXML that drives speech recognition and text-to-speech engines.
DSR Optimised Codec: The module in UE which takes speech input, extracts acoustic features and encodes them with a scheme optimised for speech recognition. This module is similar to the conventional codec, such as AMR. On the server-side, the uplink encoded stream can be directly consumed by speech engines without having to be converted to a waveform.
Meta information: Data that may be required to facilitate and enhance the server-side processing of the input speech and facilitate the dialog management in an automated voice service. These may include keypad events over-riding spoken input, notification that the UE is in hands-free mode, client-side collected information (speech/no-speech, barge-in), etc….
Speech Recognition Framework: A generic framework to distribute the audio sub-system and the speech services by sending encoded speech between the client and the server. For the uplink, it can rely on conventional (ASR) or on DSR optimised codecs where acoustic features are extracted and encoded on the terminal.
Speech Recognition Framework-based Automated Voice Service: An automated voice service utilising the speech recognition framework to distribute the speech engines from the audio sub-system. In such a case the user voice input is captured and encoded, with a conventional or a DSR optimised for speech recognition as negotiated at session initiation. The encoded speech is streamed uplink to server-side speech engines that process it. The application dialog manager generates prompts that are streamed downlink to the terminal.
SRF Call: An uninterrupted interaction of a user with an application that relies on SRF-based automated voice services.
SRF Session: Exchange of audio and meta-information, explicitly negotiated and initiated by the SRF session control protocols, between terminal (audio-sub-systems) and SRF-based automated voice services. Sessions last until explicitly terminated by the control protocols.
SRF User Agent: a process within a terminal that enables the user to select a particular SRF-based automated voice service or to enter the address of a SRF-based automated voice service. The user agent converts the user input or selection into a SIP IMS session initiation with the corresponding SRF-based automated voice service. The user agent can also terminate the session with the service when the user device to disconnect.
Text-to-Speech Synthesis: A technology to convert text in a given language into human speech in that particular language.
Uplink exchanges: Exchanges from the mobile terminal to the server / network.
For the purposes of this document the following abbreviations apply:
AMR – Adaptive Multi Rate
DSR – Distributed Speech Recognition
DTMF – Dual Tone Multi-Frequency
IETF – Internet Engineering Task Force
IMS – IP Multimedia Subsystem
IVR – Interactive Voice Response system
PCM – Pulse Coded Modulation
PIM – Personal Information Manager
SIP – Session Initiation Protocol
SRF – Speech Recognition Framework
URI – Uniform Resource Identifier