About Us

The DSP Dimension started out as a platform for online courses and tutorials on Digital Signal Processing in 1999 as part of Stephan Bernsee’s lectures on advanced audio DSP held at various institutions around Germany. The DSP Dimension quickly became a standard resource for specialized DSP topics like time stretching and pitch shifting and is part of engineering classes at several universities across Europe as well as a frequently featured content in many popular magazines. Today Stephan and his team are providing consulting and development services to the audio signal processing market and are developing applications marketed through their partners around the world.

Stephan M. Bernsee (born Stephan Sprenger) is a Signal Processing applications developer located in Mainz, Germany, working in the music/audio DSP business since 1990. Being the founder of the company Prosoniq he has led the company through many years of difficult decisions in the short lived music and audio software market. With the incorporation of Prosoniq in the UCMG network in 1993, he became senior developer of the R&D department for media and production at UCMG due to exceptional achievements. In 1995, Prosoniq GmbH was founded as a separate entity and Stephan took on a position as head of R&D. Until 2004 he was employed as Director of Research and Development at Prosoniq headquarters in Karlsruhe, Germany. Today he is working as an independent technology provider licensing his developments to numerous companies in the market, including Prosoniq.

Born in Karlsruhe, Baden, Germany in 1971 his interest in music DSP applications arose from his prior work as software consultant for a medical data processing company where he used Artificial Neural Networks for time series evaluation of medical data. He soon realized that the same technology can be successfully applied to audio processing and he subsequently developed the ANN-CASE Artificial Neural Network design tool for SGI computers that allowed parametrized development of ANN topologies and training methods on IRIX workstations.


He has since pioneered several novel methods of analyzing and rebuilding audio signals at very high quality based on new transformation methods and highly optimized neural network processing. Being the developer of Prosoniq’s unique and proprietary MCFE audio decomposition engine he has contributed the basic techniques and important later refinements to implement new audio algorithms and processing capabilities that can only be found in Prosoniq’s product line. The inventor of the MPEX near-lossless time stretching and pitch shifting, he currently works on projects on auditory scene analysis, high quality time stretching and pitch shifting and on the design and implementation of high speed neural computing and training methods.

Main work (excerpt):

Music Industry
 My main area of work has been in the music industry, although I’ve been working in other areas as well (s.b.).

1988-90

  • First model-based structural audio morphing software implemented on an Atari ST
  • Author for various music-related magazines (KEYS magazine in Germany), various articles on Digital Signal Processing and DIY courses

1990

  • Self-employed, founded Prosoniq Products Software as privately held company who delivered audio software for the ATARI 1040ST and for various embedded systems based around the Mot. 68k, 56k and custom DSP chips.

1991 onwards

  • Study of Communications Engineering in Karlsruhe
  • Co-Development of Medical diagnosis software for a client of Nemectron Corp., Germany
  • Research and specialization on adapted basis transforms and multiresolution signal analysis

1993

  • Acquisition of Prosoniq by the UCGM Corporation, Heidelberg, Germany to expand their mastering/post production dept. and to produce custom tailored DSP software development for their needs
  • Extended study and development of processes based on Artificial Neural Networks and number theoretic transforms

1993-95

  • Software developer and senior software developer at Prosoniq/UCMG, also responsible for coordination and supervision of the UCMG studios and mastering facilities and processes
  • Development of an ANN Topology Editor with C Source Code Generator for SGI IRIX 5.2-6.2
  • Development of a modular DSP compiler that generates C source code from high-level language scripts
  • Vinyl cutting simulator for optimization of vinyl cutting process for UCMG
  • Developed a psychoacoustic loudness maximizer optimzied for vinyl and CD
  • Sibilancy suppressor based on sibilance recognition technology employing ANNs
  • Pitch Shifting algorithm based on the fast Fourier transform
  • Supervised collection of a speech database and its implementation for later use in speech recognition products
  • Devised business plan and technology/component assembly for a 56k based standalone post-production device (development discontinued due to marketing reasons)
  • Discussed and planned co-development of a AD65xxx based synthesizer with a major company in the MI business
  • Devised a novel noise reduction based on adaptive transform technology and ANNs
  • Co-development of sonicWORX, a state-of-the-art audio editor
  • Development of numerous plug in modules for sound manipulation within sonicWORX
  • Developed 1/3 octave band analyzer for audio
  • Extensive work on multiresolution sinusoidal modelling using enhanced MQ and D+S signal decomposition/synthesis, extended into non-rectangular and adaptive transform based methods
  • Prosoniq GmbH was founded with Bernhard Bouché as CEO. Moved office from the UCMG facilities to the Karlsruhe Center of Technology

1996-1997

  • Development of a real time lapped transform filterbank for speech coding and vocoder products (later to become part of OrangeVocoder)
  • Development of a real time LOT resynthesizer (later to become part of Prosoniq’s Magenta product)
  • Extensive modelling of analogue circuitry (later to be used in the products OrangeVocoder and NorthPole)
  • Devised novel multiresolution based formant manipulation to allow for natural sounding monophonic and polyphonic pitch changing (later to be used in Emagic’s Logic Audio range of products)
  • Developed a set of realtime effects to complement Emagic Logic Audio 3.0
  • Development of a fast shift invariant wavelet transform for audio coding and analysis
  • Development of MCFE time compression/expansion based on complex-valued wavelet packet transform

1997-1999

  • Developed acoustic 3D scenery modelling software based on a cellular room modelling approach (later to be implemented into Prosoniqs “Ambisone” product)
  • Coordinated porting and feature set of Prosoniq’s VST products on the Digidesign RTAS platform
  • Coordinated and planned feature set development and implementation for Hartmann Neuron Synthesizer
  • Co-developed “NorthPole” with Frederic Schelling. NorthPole is a 4pole resonant filter for the VST interface, to date still the most widely distributed free plug in for VST
  • Developed realtime lead vocal reduction software based on adaptive, model-driven transform technology, later implemented in Pandora Realtime for sonicWORX
  • Commenced work on forensic voice separation and blind signal separation software for various clients and institutions, work in progress
  • Developed new dithering process for requantization based on geometric series to allow for minimum noise and maximum clarity (later to be used in sonicWORX)
  • Development of a time scaling technique based on the fractional Fourier transform
  • Developed time compression/expansion technology for NewMotion GmbH
  • Developed custom tailored DSP and filtering applications for NewMotion GmbH
  • Developed time compression/expansion technology for DAVE GmbH
  • Frequent poster to comp.dsp, also involved in DSP training and education in Karlsruhe, Heidelberg and Darmstadt

1999-2002

  • Developed specialized custom-tailored sample manipulation tools for YellowTools makers of sampling CDs, most notably an automatic pitch-synchronous loop finder
  • Developed method for loop-matching the contents of a sampled sound to allow for smooth loops within a user-defined area
  • Developed Prosoniq’s MPEX and MPEX2 (multi-channel) near-lossless time compression/expansion technology which was highly acclaimed by the public and was awarded several prices. Today it is an integral part of all major music production software packages.
  • Implemented MPEX technology for watermark recovery for BlueSpike techology
  • Implemented MPEX Time compression/expansion technology for Steinberg Nuendo
  • Implemented MPEX Technology for Prosoniq TimeFactory which was awarded several prices
  • Implemented MPEX technology for IRCAM/Apple software
  • Carried out optimization and parallelization of all Prosoniq algorithms on G4 processors (Altivec technology) and processor specific optimizations for Pentium III processors
  • Developed a multiresolution realtime resynthesis technology for sound reproduction for up to 56 voice polyphonic realtime sinusoidal resynthesis on a G4 processor
  • Implemented enhanced phase coherence pvoc for Roni Music, Sweden.

2002-present

  • Extensive research into sound modelling and parametrization based on adaptive transforms and artificial neural networks
  • Developed real-time polyphonic rendering engine based on structural audio information
  • Developed analysis library to perform parametrization and model extraction on sampled sounds
  • Conducted extensive studies on acoustic performances and instrument timbres to gather data for parametrization training
  • Developed NAS technology and custom-fitted it to the Hartmann <Neuron> synthesizer
  • Adapted MPEX2 time compression/expansion technology for use in Steinbergs Cubase SX
  • Adapted MPEX2 time compression/expansion technology for use in Merging Technology’s Pyramix system
  • Development of the “Prosoniq morph” realtime audio morphing plug in for VST/RTAS
  • Consultation work for several major companies in the MI and other businesses
  • Development of the Neuron synthesis technology used in the Hartmann Music range of products

2002-2003


 Optics

  • Developed a wavelet based image decomposition tool to allow for optically pleasing image scaling and transformation
  • Developed an image filtering toolbox based on local fractional Fourier transforms to enhance research data collected by a LCD sensor
  • Developed real-time vision-aided onset detection and phoneme recognition software to aid in collecting data for a TTS database

Welcome

Welcome to The DSP Dimension, the internet platform for our engineering training and software development services in the audio and video applications market. My team and I, Stephan Bernsee, specialize in advanced signal processing for audio and video applications on general purpose desktop computer systems and dedicated DSP hardware. We bring 20 years of development experie […]