Manipulation of relevant input features

• LocCNN

  

• SampleCNN

  

Feature Inspection

In the following, for each environmental condition considered we will report feature results related to the best and worst performing sources for both LocCNN and SampleCNN

T60 0.15 s SNR 25 dB

SampleCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

LocCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

T60 0.3 s SNR 20 dB

SampleCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

LocCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

T60 0.4 s SNR 15 dB

SampleCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

LocCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

T60 0.6 s SNR 10 dB

SampleCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal

LocCNN

• Best performing

  

Microphone audio signal

Relevance audio signal

• Worst performing

  

Microphone audio signal

Relevance audio signal