OSP Release 2018A

WHAT IS NEW IN OPEN SPEECH PLATFORM 2018A?

  1. Release 2018a introduces baseline version of an advanced Hearing Aid feature, namely Speech Enhancement (SE) for improving intelligibility in noisy environments. The baseline SE consists of (i) Peak and Valley detection, (ii) Noise estimation and (iii) Spectral Subtraction using Wiener Filtering. Three approaches for noise estimation are provided — (a) Arslan Averaging Procedure, (b) Hirsch & Ehrlicher weight averaging procedure and (c) Cohen & Berdugo MCRA Procedure. The baseline SE is provided in source code.
  2. A new AFC module based on “Sparsity Promoting LMS (SLMS)” is added in compiled form to OSP. See here for a presentation on SLMS.
  3. Support for input 48 kHz has been added to OSP. Some newer MacBooks support only 48 kHz sampling for the internal codec. Researchers can connect an off-the-shelf headset with a 3.5 mm (1/8”) mini stereo plug and test the  Real-Time Master Hearing Aid (RT-MHA). Choose 96 kHz option in Apple Menu - System Preferences and Input, if your MacBook supports higher smiling frequency for the internal codec.
  4. For the researchers investigating HA processing with ear-level assemblies, we have developed behind the ear, receiver in the canal (BTE-RIC) form-factor-accurate hearing aids (we call them Jelly Beans or JBs) that connect to a break out box (BoB), which in turn connects to an audio interface box with thunderbolt interface to the MacBook. We have a limited supply of these and we provide them to our audiology and speech science collaborators. if you are interested. We have included use of JBs and BoB in the documentation, along with other improvements.

NEWS + RELEASES: