IEEE Recommended Practice for Speech Quality Measurements

Abstract

The increasing variety of speech transmission systems has created a new problem for communication engineers. This problem consists of the subjective evaluation of the speech quality produced by these systems. For investigators interested only in the utilization of some procedure applicable to this problem, it was difficult to make a choice among the wide range of documented approaches. The IEEE Audio and Electroacoustics Group therefore formed a Subcommittee on Subjective Measurements, which has been charged with writing a Recommended Practice for speech quality measurement. The intent of the Subcommittee became to write a document that would facilitate the choice of method by reducing the variety of procedures to several which had been successful and looked promising for the future, and which could handle a variety of signals. By describing these methods in some detail, it would be possible for different laboratories to obtain comparable results and thus aid in the exchange of information. Because of the relatively early stage of the development of speech quality evaluation, the present document is intended to be a Recommended Practice, but not a long term standard, which would require speech quality measurement only according to the proposed methods. It is hoped that the document will generate additional investigation into problems of this type of measurement, and allow comparison of test data obtained in different laboratories. Thus, a body of experimental data may be generated that will be useful to future investigations as well as the improvement of the present practices. The Subcommittee expects that feedback from users of the practices will allow the revision and updating of the practices within a relatively short time period. It should be pointed out that the present practices evaluate only speech signals as they are generated by one-way speech communication systems. Systems, especially two-way systems, in which interactive communication takes place, cannot validly be evaluated by these methods. One of the directions, therefore, in which these practices might be updated in the future is by adding material on the evaluation of communications systems. It is the hope of the Subcommittee that the present document will be helpful in some way to people confronted with the task of making speech quality measurements. In order to guide its future work, the Subcommittee would appreciate that any technical comments about the usefulness of the documents, or any other comments or information about speech quality measurement, be sent to its Chairman, c/o the IEEE Audio and Electroacoustics Group, IEEE Headquarters.

Scope

Speech quality is based on the subjective appraisal of speech. Speech may be appraised on the basis of preference, loudness, intelligibility, and recognizability of properties of the original speaker's voice, of the transmission channel, and/or in the case of artificial speech, of the speech synthesizer. The IEEE Subcommittee on Subjective Measurements, charged with writing an engineering practice for the measurement of speech quality, has concluded that a single method should not now be recommended. The present Recommended Practice is concerned only with preference measurements for which three methods will be tentatively outlined. These are the Isopreference Method, the Relative Preference Method, and the Category-Judgment Method.

Topic

Signal Processing and Analysis, Communication, Networking and Broadcast Technologies