LOSSY AUDIO COMPRESSION BASED ON CURVE INTERPOLATION AND DYNAMIC PROGRAMMING

V diplomski nalogi pregledamo področje stiskanja avdio posnetkov in razvijemo lasten algoritem za stiskanje avdio posnetkov v valovni obliki oz. v časovnem prostoru. Algoritem temelji na zamenjavi posameznih vzorcev znotraj avdio posnetka z daljico ali z Bézierjevo krivuljo drugega reda. Za optimalno izbiro nadomeščenih odsekov se uporablja dinamični algoritem veriženja eksonov, ki ga povzemamo po napovedovanju genov v bioinformatiki.In this work we examine the area of audio compression and develop our own algorithm for compressing audio recordings in waveform, ie. time domain. The algorithm is based on replacing individual samples in the audio recording with lines or second order Bézier curves. For choosing optimal replacement segments a dynamic programming algorithm of exon chaining is used, which is modelled after gene prediction in bioinformatics