Loudness matching for compressed speech signals
Matthew H. Bakke, Arlene C. Neuman, Harry N. Levitt · The Journal of the Acoustical Society of America · 1991
The purpose of this experiment was to determine a rule for matching the loudness of uncompressed and compressed speech samples. Eight normal hearing and 16 sensorineural hearing-impaired subjects (divided into four groups of differing severity and configuration) listened to two continuous speech samples (male and female talkers) under conditions of linear and compression amplification. Compression conditions consisted of 12 selected combinations of varying compression ratio, threshold, and release time (compression ratio = 2:1, 4:1, 8:1; knee point = 5, 10, 15, and 20 dB below the highest speech peak; release time = 20, 200 ms). Subjects matched the loudness of the compressed signals to a reference uncompressed signal using an adaptive procedure. Overall levels and cumulative distributions of the processed speech signals were obtained. Estimates of equal loudness based upon rms level and 90th percentile of the cumulative distribution were both found to closely match the gain selected by subjects. [Work supported by NIH Grant No. 2PO1 DC00178.]