Bell Atlantic Network Services, Inc.

Verizon Services Corporation

James R Schrage

Recording and automated transcription methods and apparatus suitable for use in a communication system such as a telephone system are described. In one embodiment, a telephone system has a conference bridge, a transcription system, and multiple telephone sets connected to a central office (CO) switch via communication channels. Each telephone set lets its user transmit speech on first and second channels simultaneously. The CO switch selectively connects the first channel from each telephone being used in a conference call to the conference bridge. The bridge interconnects the first channels to establish a telephone conference. The CO switch selectively connects the second channels to the transcription system, which records the teleconference participants in separate recordings. The separate recordings are each time stamped and identified with a user's or communication unit ID. An automated speech-to-text (speech recognition) system transforms the recorded speech into textual data. A collator collates text segments generated from the speech obtained from different channels using the time stamps and the ID's to form a master transcript of textual data. The present invention can be used by single or multiple users. A single user may use the system of the present invention as a dictation system. The system can be used for automatically generating a transcript of a multi-party telephone conference or meeting.

Methods and apparatus for providing speech recording and speech transcription services

Hands free voice dialing telephony devices that can perform relatively simple speech recognition, e.g., to recognize one or a few words corresponding to a command to initiate voice dialing, are described. Speech recognition models stored in the telephony devices can be relatively small and may be either of a speaker dependent or speaker independent type. In response to detecting a command to perform a voice dialing operation the telephony device establishes a connection with a voice dialing intelligent peripheral (IP). The IP includes far greater speech recognition capabilities than the individual telephone devices and is responsible for supporting voice dialing operations associated with a plurality of voice dialing service subscribers. The IP performs speech recognition on speech provided by individual telephony devices and outputs telephone numbers corresponding to recognized spoken names. Telephony devices are coupled by the telephone network to destination telephones corresponding to the telephone numbers output by the IP. In one embodiment, speech recognition models are generated by the IP from speech transmitted from the individual telephony devices. The generated model or models are then stored in the telephony devices for use during speech recognition operations. Thus, processing resources required to generate speech recognition models can be located in a centralized network accessible location.

Growing community of inventors

Ridgefield, CT, United States of America

James R Schrage