Temporal envelope and fine structure cues for speech intelligibility

Abstract

A number of listening experiments were run to investigate the relative contribution of temporal envelope modulations and fine structure to speech intelligibility. Firstly, the amplitude envelopes of 24 1/4-oct bands (100–6400 Hz) were processed in several ways (e.g., fast compression) to assess the importance of modulation peaks and troughs. Results show that reduction of modulations by the addition of noise is more detrimental to sentence intelligibility than the same degree of reduction achieved by direct manipulation of the envelope; in some cases the difference in speech-reception threshold (SRT) is almost 10 dB. Here, Leq−3 dB defines a perceptual cross-over level for which removing modulations either x dB below or above yields the same intelligibility score, whereas Leq−15 dB defines the median envelope level (acoustic cross-over level). Secondly, the fine structure of each 1/4-oct band was altered by manipulating the distribution of speech and noise over the sentences. Results indicate that noiseless peaks do not increase intelligibility; removing the speech fine structure in the troughs, however, yields a 2-dB increase of the SRT. In general, no one-to-one relation between the modulation-transfer function (MTF) and the intelligibility scores could be established. [Work supported by NWO.]

Effects of directional microphone and adaptive multichannel noise reduction algorithm on cochlear implant performance

“…These results were also consistent with findings of studies involving listeners with acoustic hearing ͑Steeneken and Houtgast, 1982;Payton et al, 1994;Goldworthy, 2005;Ricketts and Hornsby, 2003͒. In contrast, the ranges of calculated STIs were 0.0129 for ON and 0.0082 for DN across the SNRs while the ranges of the speech recognition scores varied between 40.1% and 77.8% for the three groups of listeners. These results in electrical hearing parallel the findings of previous studies in acoustic hearing that STI is a poor predictor of speech recognition scores for nonlinearly processed speech ͑Ludvigsen et Drullman, 1995;Hohmann and Kollmeier, 1995;Goldsworthy, 2005͒. The extremely narrow range of STIs suggests that the modulation depth of the speech temporal envelope is almost identical across the SNRs.…”

“…In general, the higher the modulation depth in the transmitted or processed signal, the higher the speech transmission index and the higher the predicted speech intelligibility. The disadvantage is that STI has been reported to fail to predict the speech intelligibility of nonlinearly processed speech for people using acoustic hearing ͑e.g., for spectral subtraction noise reduction algorithm, see Ludvigsen et al, 1993, Goldsworthy andGreenberg, 2004; for envelope clipping, see Drullman, 1995;for compression, see Hohmann and Kollmeier, 1995; for envelope thresholding, see Goldsworthy, 2005͒. It appeared that factors other than the modulation depth affected the performance of people using acoustic hearing.…”

Probabilistic Encoding of Vocalizations in Macaque Ventral Lateral Prefrontal Cortex

“…We also found that the performance of the HMM was reasonable with between four and 10 spectral channels, but using only two or Ͼ10 channels led to a decrease in performance. These results are consistent with studies of speech perception that have demonstrated relatively high quality speech recognition with only a few spectral channels (Drullman et al, 1994a,b;Drullman, 1995;Shannon et al, 1995). In principal, we could also filter the temporal dimension in the spectrogram, before fitting the HMM, to determine how much temporal resolution is necessary for discriminating among the calls.…”

Transformation of Temporal Properties between Auditory Midbrain and Cortex in the Awake Mongolian Gerbil

“…Interestingly, they sacrifice some precision in representing acoustic fine structure to reliably encode the sound envelope. This is well matched by the fact that speech perception is quite robust to disturbances in fine structure but depends critically on the amplitude envelope (Drullman, 1995;Shannon et al, 1995). Although it is known that temporal fine structure is important in some speech reception situations such as in noisy environments and for tonal languages (Zeng et al, 2005;Kong and Zeng, 2006), our study only addresses envelope processing.…”