Syntronic AB is a global consultant company with 650 employees. The company is based in Gävle with offices in Göteborg, Kista, Linköping, Sandviken, Beijing, Chengdu, Ottawa and Kuala Lumpur. Our expertise in Göteborg is primarily in embedded systems, data science and algorithms. We seek two talented and dedicated master students with strong programming skills, and some knowledge in signal processing and machine learning.
Real time speaker recognition
In automatic speech-to-text transcription there are several non-semantic factors such as age, gender, dialect, emotion, speaker that can degrade the performance of the transcription if not taken into account. On the other hand, if classification of such factors is done prior to transcription and the information is used in the transcription performance has good chances to improve.
A previous master thesis project at Syntronic concerned Swedish dialect classification offline. In this thesis project, methods from that thesis are supposed to be modified and evaluated online, on, e.g., a Raspberry Pi with a microphone and possibly together with a cloud solution. Emphasis will be on the implementation on the embedded hardware, local pre-processing of speech and communication with a cloud if necessary. To simplify the classification problem and allow easy annotation of recordings, speaker recognition is suggested.
The successful students will have good chances to become employees at Syntronic after a completed thesis.
Supervisor at Syntronic: Adam Andersson, email@example.com.
Please, submit a personal letter, CV and transcripts with your application.