John Usher's Homepage

Projects:

Loudspeaker design: Carbon fibre active loudspeaker

Loudspeaker design

Spatial audio pickup for beamforming:

Spatial audio pickup

Hearables and TWS Truly Wireless smart earphone design:

Hearables and TWS

Various fun projects

Control of lights using voice with Whisper-API and Chat-GPT API to generate Arduino code:

Control of lights using voice

Teaching kids to spell. A "magic" wand with IMU and GPS can detect the letter being spelt, using a custom CNN neural-net

Teaching kids to spell

Determining a guitar chord using a single microphone with low-latency (<20 ms), ~90% reliable. This was a supervised learning algorithm that uses a generative model, specifically Gaussian Mixture Models (GMM) trained on Mel-Frequency Cepstral Coefficients (MFCC) features, for audio classification.

Science-fiction writing:

Science-fiction writing
Experience:

John Usher is a humble (ish) pioneer of spatial audio enhancement algorithms for earphones and loudspeakers. He holds a B.Eng. in Electro-Acoustics from the University of Salford and a Ph.D. from McGill University, where he was advised by such Masters as: Dr. W.L. Martens; Prof. W. Woszczyk; Dr. J. Benesty and Prof. A. Bregman. The Ph.D. introduced and characterised a new loudspeaker audio upmixing system, further developed for multichannel cinema audio at IMM sound, and now part of Dolby Atmos (patent #8,335,330).

Dr Usher has conducted research at the acoustics department of Bang and Olufsen, the Audio DSP group at Philips Nat Lab, and has given talks at Google-X, University of Cambridge, McGill University, and NASA.

John is an expert and pioneer in the field of Hearables: wearable electronic devices relating to sound. In 2006 he joined a new company designing and building smart Hearables. At Techiya (originally Personics, then Techiya), he spent over 10 years conjuring a new earphone for hearing enhancement, augmented reality, hearing protection (dosimetry), music listening and voice communications: it was, arguably, the world's first smart consumer earphone.

John then co-founded a loudspeaker company. He engineered an all-carbon fibre high-end active loudspeaker. The design used a novel waveguide to direct sound towards a listening position, reducing coloration from room effects. The sound experience was commended by illustrious industry critics and musicians.

He now works in the audio group at Synaptics, engineering next-generation telecommunications systems using classical DSP and the latest ML systems. He is a keen tinkerer of electronics projects using Arduino and Raspberry Pi.

Specialties:

Audio enhancement with DSP and ML for noise reduction, dereverberation, AEC, automatic level control, spatial filtering. Audio ML using RNN, CNN, U-Nets in Pytorch and Tensorflow with Onnx export. Psychoacoustics for loudspeakers and earphones: timbral, spatial fidelity optimization, situational awareness (passthrough). Low-latency audio DSP algorithms: ANC, echo-cancelling, beamforming, filtering, sound detection/ classification, sub-band processing, noise reduction, spatial filtering. Dosimetry. Room acoustics. Spatial audio upmixing. Subjective sound quality evaluation for music and speech. Code: Matlab, C, Python, assembly, Go. Patent drafting: 50+ patents on audio DSP. Excellent communication and independent project management skills. Leet Prompt Mashing.