ML-KWS-for-MCU

ML-KWS-for-MCU copied to clipboard

When I try to run my own audio recordings using label_wav.py, I always get random predictions. I trimmed the audio files to only contain the keyword and still receive false outputs. Does anyone experience the same?

Same experience here at first, but I probably solved it by checking the format of my own wav file. Your wav file should have the same format as those in training data. i.e., the sample rate is 22050 and bit rate is 256kb/s, the length of the audio must be exactly 1 second. You can check the details by looking at Google speech command dataset.