Informed Spectral Analysis: improving audio quality using estimation and information


Informed Spectral Analysis is a two step approach for spectral analysis that consists in combining a classic blind estimator with the extra-information necessary to reach the desired precision.

Motivation:
  • music content creator have access to separated tracks before the mixing process
  • Blind analysis method have theoretical limitation for the best reachable precision
    		•

    Proposal of solution:
    a novel method based on a coder / decoder configuration which inaudibly embed the minimal necessary extra information required by by the chosen estimator to obtain the desired quality.

    Coder:

    Decoder:

    Practical experiment: For this experiment we consider a single-channel 44.1kHz-sampled music signal with 6 sources. At the coder, sinusoidal parameters are estimated and the required extra information to reach the target precision using the reassignment method from the mixture is computed and inaudibly embedded into the mixture. At the decoder, the extra information is extracted from the mixture signal and combined with the reassignment method to recover signal parameters. For this experiment, we use the Hann analysis window of length N=1023 with 50% overlap.
    Sound results:
    the proposed audio were obtained for a realistic mixture using a overall bitrate of 147.76 kbps for the informed method corresponding to a theoretical watermark capacity of 155.65 kbps. The average bitrate per source is 24.62 kbps. The total bitrate to reach the same audio quality using ECUSQ is 249.60 kbps.
    Source 1: guitarSource 2: bass
    blind estimation - informed estimation - original blind estimation - informed estimation - original
    Source 3: drumSource 4: synthesizer
    blind estimation - informed estimation - original blind estimation - informed estimation - original
    Source 5: voiceSource 6: guitar 2
    blind estimation - informed estimation - original blind estimation - informed estimation - original


    Overall bitrateAverage bitrate per source

    Original Mixture Watermarked mixture Semi-blind estimated mixture Informed estimated mixture
    wav wav wav wav