Testing the ability of speech recognizers to measure the effectiveness of encoding algorithms for digital speech transmission

Modern communication channels, such as digital cellular telephony, often convey human speech in a highly encoded form. Methods that rely on human subjects to evaluate the quality of such channels are too costly to deploy on a large scale; thus, automated methods are often used to model quality as perceived by humans. Traditional automated methods that use signal to noise ratios (SNR) to judge the quality of channels do not model human perception well when applied to highly encoded speech. For this reason, researchers investigate alternative means to objectively measure the quality of such channels. We explore the feasibility and applicability of using automated speech recognition technology to model human perception of the quality of communication channels that carry highly encoded (compressed) human speech.