Sound Demo for the Tandem Algorithm for Speech Segregation

This page demonstrates the tandem algorithm proposed by G. Hu and D.L. Wang. For details of this algorithm see:

Hu G. and Wang D.L. (2008): A tandem algorithm for pitch estimation and voiced speech segregation. IEEE Transactions on Audio, Speech, and Language Processing, vol. 18, pp. 2067-2079.
For the second set of demos with a naturalic utterance (including both voiced and unvoiced speech), an unvoiced speech segregation algorithm is additionally used. This unvoiced segregation algorithm is described in:
Hu G. and Wang D.L. (2008): Segregation of unvoiced speech from nonspeech interference. Journal of the Acoustical Society of America, vol. 124, pp. 1306-1319.


Voiced speech segregation (for comparison with earlier systems click here)

    Noise (mixture ID)    Mixture    Segregated target 
Pure Tone (v3n0)
White Noise (v3n1)
Noise Burst (v3n2)
Cocktail Party (v3n3)
Rock Music (v3n4)
Siren (v3n5)
    Trill Telephone (v3n6) 
Female Speech (v3n7)
Male Speech (v3n8)
Female Speech (v3n9)


Naturalistic speech segregation (SNR = 0 dB)

    Noise    Mixture    Segregated speech 
White Noise
Rock Music
Electric Fan
Alarm Clock
  Bird Chirp with Water Flow  
Wind Noise
Rain
Cocktail Party
Playground
Crowd Noise