Speaker-specific speech-to-text/text-to-speech communication system with

Abstract

The invention consists of speech compression and transmission devices (CTD) connected to one another for transmission of signals therebetween via any transmission media including the public switched telephone network, a cellular network, a satellite or submarine communications system or any combination of such existing networks and systems. Each CTD consists of a speech profile defining the coefficients of speech for any person that will be using that device, a means for converting speech to text, and a means for converting text to speech based on the speech profile. When a connection is established between the devices, the speech profiles of the calling party and the called parties are exchanged such that the calling party's CTD stores in RAM the called party's speech profile and the called party's CTD stores in RAM the calling party's speech profile. When either one of the parties speaks, that party's speech is converted into text. Changes in picth, rate, and volume of the speech are detected, and the amount of change indicated by hypertext characters inserted in the text. The text is transmitted from the speaking party's device to the listening party's device where it is converted back into speech using the speaking party's speech profile. Because speech typically occurs at a rate of between 4 and 5 words per second with an average of 4.2 characters per word, normal speech generates approximately 20 characters per second. Assuming the formatted text uses 8 bits to define a character such as for standard ASCII text plus hypertext characters, the system of the invention transmit speech at a data rate of 160-300 bits per second, a data rate significantly lower than known speech compression techniques.