Patent US 4,170,719

A speech transmission system is improved in intelligibility and naturalness by separating voiced from invoiced speech segments prior to application to a transmission channel of restricted bandwidth. Voiced segments are combined without processing with discrete-frequency coded unvoiced segments processed in analog or digital fashion conformably with the limited channel bandwidth at the transmitter. Voiced segments are reproduced conventionally while unvoiced segments are simulated by noise sources triggered by decoded discrete frequencies at the receiver. The reconstructed speech signal can thus occupy substantially more than the limited channel bandwidth.