The I/O relations in correlation vectors of speech are (3).
The K-dimensional MP coefficient vectors of speech, noise, and noisy speech are denoted as α s, α n, and α with their k th elements α s,k, α n,k, and α k, respectively.
Codebook-driven speech enhancement techniques [25],[26] estimate m x and m w for each short-time segment: a x and a w are selected from trained codebooks of vectors of speech and noise LP coefficients, C x and C w, respectively, and the gain terms g x and g w are computed online, resulting in good performance in nonstationary noise.
The quantization scheme of [18] is followed here, which was developed for vector quantization of speech line spectral frequencies (LSFs).
MFCCs and their first and second derivatives were used for the 39-dimensional vector representation of speech signal every 10 ms. A bigram language model was used for decoding.
A state-of-the-art speaker verification system is based on the low-dimensional i-vector representation of speech recordings.
Poria et al. [15] built higher level features from textual data, represented by a 306 dimensional vector consisting of a word vector and part of speech values, by training a CNN.
However, the envelope of compressed data still retains considerable information of the original speech [20, 21], we therefore alias and scramble the two measurement vectors of separated speech instead of using the envelope as ciphered data directly.
GMMs are suitable for this problem since they have been shown to model successfully the statistics of spectral vectors of both speech [18, 23] and audio signals [19].
Our basis features are C-dimensional Mel-spectral feature vectors of observed speech y that have been normalized to compensate for spectral distortion.
