Cochlear-Implant Simulations

A noise-excited vocoder simulates certain aspects of cochlear-implant processing, by reducing spectral information and discarding temporal fine-structure information. This has surprisingly little effect on speech intelligibility in quiet:[Note: Sound files will require appropriate browser plugins]

1. Unprocessed sentence:
   
2. Processed with a 24-channel noise-excited vocoder:
   

However, in more complex situations, where the speech has to be segregated from competing sounds, cochlear-implant processing has a highly detrimental effect on intelligibility.

3. Unprocessed sentence in background speech:
   
4. Processed with a 24-channel vocoder:
   

For more details see Qin and Oxenham (2003).

Auditory Stream Segregation

A sequence of tones alternating between two frequencies, A and B, like ABAB….. , can give rise to completely different percepts, depending on the A-B frequency separation Df:

• If Df is small (i.e., 1 semitone, about 6%), the sequence is usually perceived as a single stream of sounds alternating in frequency
• If Df is large (e.g., 9 semitones), the sensation of pitch alternation is lost and most listeners perceive, instead, two streams of constant-frequency tones running in parallel.

The difference in percept is even more striking if the tones are grouped into triplets (ABA…ABA…), as the sequence then evokes a distinctive “gallop” rhythm at small-?fs, but for large ?fs, the gallop is gone.

At intermediate delta-Fs, the sequence is usually perceived as a single stream (gallop) at first, but often splits into two streams after a few seconds of uninterrupted listening; this phenomenon is known as the build-up of stream segregation.

1. Galloping sound at small Df (Single stream) :
   
2. Gallop vanishing for larger Df (Double stream) :
   
3. Intermediate Df :
   

Auditory Chimeras
ASA Sound Examples