Robust speech recognition using the modulation spectrogram

By Brian Kingsbury, Nelson Morgan, Steven Greenberg. 1998

Subfields: Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Speech & Audio

In Signal Processing

Citations in the corpus

Robust wav2vec 2.0: Analyzing Domain Shift in Self-Supervised Pre-Training (2021, background): “Both signal processing-based [11] and learned [12] features have been explored.”
A Spectral Energy Distance for Parallel Speech Synthesis (2020, neutral): “Following the literature on speech recognition [20, 1], we thus define our distance function over spectrograms s(xi), where a spectrogram is defined as the magnitude component of the short-time Fourier transform (STFT) of an input waveform, |STFTk(xi)|, where k is the frame-length used in the STFT.”
Extracting Domain Invariant Features by Unsupervised Learning for Robust Automatic Speech Recognition (2018, background): “Alternatively, robustness can also be achieved by training using features that are domain invariant [10, 11, 12, 13, 14].”
Automatic speech recognition and speech variability: A review (2007, methodology): “Some representative examples are temporal patterns (TRAPs) features (Hermansky and Sharma, 1998), MLPs and several modulation spectrum related techniques (Kingsbury et al., 1998; Milner, 1996; Tyagi et al., 2003; Zhu and Alwan, 2000).”
Speaker Diarization: A Review of Recent Research (2010, methodology): “The main approach consists of a three state HMM-GMM ystem (non-speech, non-overlapped speech, and overlappe d speech), and the best feature combination is MFCC and modulation spectrogram features [103], although comparab le results were achieved with other features such as root mean squared energy, ”
Size matters: an empirical study of neural network training for large vocabulary continuous speech recognition (1999, methodology): “As part of our Broadcast News e ort, we are also employing a di erent set of 28 features based on the modulation-spectrum, using a modi ed form of the approach described in [6].”
Continuous latent variable models for dimensionality reduction and sequential data reconstruction (2001, background): “The most common variants are the mel cepstrum4 (MFCC) and perceptual linear prediction (PLP) (Hermansky, 1990), which provide very similar features; a more recent proposal are modulation-filtered spectrogram (MSG) features (Kingsbury et al., 1998), developed for speech recognition robust to acoustic”
Sparse Multilayer Perceptron for Phoneme Recognition (2012, background): “Depending on the manner in which features are derived, they can be broadly classified into spectral [17], [31], temporal [32], [33], or spectro-temporal [34]–[36] features.”