Sentence examples for enhancement tasks from inspiring English sources

This mapping approach is applied to both speech enhancement and ASR feature enhancement tasks.

The software design is aimed at the algorithmic-level decrease of the computational load of the large-scale SAR image enhancement tasks.

We address the problem of acoustic source separation in a deep learning framework we call "deep clustering". Deep learning has recently produced major improvements in speech enhancement tasks in which the speech and interference belong to distinct classes of signal.

In addition, from the perspective of standardization bodies, MTC enhancement tasks are ongoing in 3GPP Release 12 and upcoming Release 13 for optimizing mobile networks to serve MTC traffic.

Three enhanced signals are produced for the speech enhancement task, and two enhanced features are produced for the ASR task.

For the speech enhancement task, our experiments show that HDNN significantly outperforms DNN in terms of perceptual evaluation of speech quality (PESQ), which is an objective measure highly correlated to subjective testing of listening quality.

The DNN-based speech coefficients mapping is applied to both log-magnitude spectrogram enhancement for speech enhancement task, and also MFCC enhancement for ASR task.

There are three stages of processing, i.e., (1) the beamforming, (2) log-magnitude spectrum enhancement for the speech enhancement task, and (3) speech features (such as MFCC) enhancement for the ASR task.

In this study, we use DNN to estimate both the low-dimensional speech feature vectors for the ASR task and a high-dimensional log-magnitude spectrum vector for the speech enhancement task.

For the speech enhancement task, the proposed dynamic feature constraint help to improve cepstral distance, frequency-weighted segmental signal-to-noise ratio (SNR), and log likelihood ratio metrics while moderately degrades the speech-to-reverberation modulation energy ratio.

In the first stage, for the speech enhancement task, we conducted experiments to understand whether utterance-wise mean normalisation should be applied to log-magnitude spectrogram to reduce channel effects on the data.