• Content-based retrieval

    Our Audio Content-Based Retrieval demos allow sounds to be compared using various features. Our software can analyze sound files, cluster them according to categories, and search for sounds that are similar or dissimilar to other sounds.

    Note: Our old demos are currently unavailable. If you wish to see try the code, you can download a copy of SoundFisher at

  • Content-based discrimination

    Our sound analyses can also be used to discriminate between broad sound classes. Download a Windows NT demo here. It listens to the PC sound input and discriminates between clean speech and music (which may include speech). It has been trained on radio recordings, so it will work best if you hook it up to a similar broadcast source or a CD player. The included training file probably won't discriminate between one person singing and talking into a PC microphone.

    The zip file contains two files: the demo and a the classifier data file. Unzip these two files in a directory of their own and double-click on the executable. Use whatever mixer control you have (typically Start->Programs->Accessories->MultiMedia->Volume Control) to make sure that you have a reasonable input level (distorted or silent inputs will probably be misclassified) and that the line input (or whatever you are using) is switched on.

  • DSP algorithms

    Our Time Scaling and Pitch Shifting algorithms allow audio files to be modified in time and pitch independently of each other. We have time-domain (fast) and frequency domain (for large changes) versions. Applications include digital music processing, voice-mail processing, and multimedia conversions.