CPC G10L 21/0208 (2013.01) [G10L 25/30 (2013.01)] | 2 Claims |
1. A system for noise reduction in aircraft simulator sounds, comprising:
an acquisition module configured to acquire sound data from an aircraft simulator sound system;
a classification module configured to classify the sound data to obtain classified audio data;
a Short-Time Fourier Transform (STFT) module configured to perform STFT processing on the classified audio data to obtain spectral frames;
a time series feature analysis unit configured to process the spectral frames by using a recurrent neural network, and extract time series features from an input noisy audio signal to capture time-dependence and dynamic changes in audio data, thereby obtaining the time series features of the audio data;
a noise reduction unit configured to input the time series features of the audio data into a deep Q-network, wherein the deep Q-network uses a deep symbolic regression algorithm to optimize processed spectral frames based on evaluation indicators comprising signal-to-noise ratio, speech distortion, and channel distortion by adaptively adjusting a window function type, a window length, a Fast Fourier Transform (FFT) length, and a hop length, thereby achieving efficient noise reduction; and
an inverse STFT module configured to perform inverse STFT on the processed spectral frames to obtain denoised audio.
|