Speech compression, or reducing the amount of data required to represent speech, is becoming increasingly vital for systems that involve speech processing. Speech compression has two primary benefits: in applications that require recording and/or playback of speech messages (such as answering machines and verbal security alarms), it helps to reduce storage requirements, enabling longer speech segments to be recorded. In applications requiring speech communications (such as walkie-talkies), compression reduces the amount of data to be transmitted, thereby more efficiently utilizing the available communications bandwidth. However, the most commonly used speech- compression techniques, especially those used for high compression ratios, are based upon standardized algorithms that typically involve significant royalty costs.

This paper describes a relatively inexpensive speech-compression technique, based upon an open-source algorithm called Speex. The Speex algorithm provides a high compression factor of 16:1, while at the same time maintaining high speech quality.