Speech-Noise Chimeras
Home Up Why Study Chimeras? Chimera Synthesis Further Information

 

    Speech-Noise chimeras are formed by pairing a speech utterance with noise having the same power spectrum.  Such chimeras are synthesized with varying number of frequency bands (see Synthesis).  Listener's performance in speech reception improves with increasing number of frequency bands for speech-noise chimeras having the speech envelope and the noise fine structure.  Consistent with the results of Shannon et al. (1995), these chimeras are intelligible with as few as 3-4 frequency bands.  In contrast, performance degrades with increasing number of bands for chimeras having the speech fine structure.  For 2-6 bands, both envelope and fine structure contain significant information for speech reception.  This condition corresponds roughly to the effective number of frequency channels in cochlear implants.

 

Speech Envelope

Speech Fine Structure

1   Frequency Band

1   Frequency Band

2   Frequency Bands

2   Frequency Bands

3   Frequency Bands

3   Frequency Bands

4   Frequency Bands

4   Frequency Bands

6   Frequency Bands

6   Frequency Bands

8   Frequency Bands

8   Frequency Bands

16  Frequency Bands

16  Frequency Bands

 
Shannon, R.V., Zeng, F.-G., Kamath, V., Wygonski, J., and Ekelid, M. Speech recognition with primarily temporal cues.  Science 270: 303-304, 1995.