T. Ananthapadmanabha

Jerrin Thomas Panachakel, A. G. Ramakrishnan, T. V. Ananthapadmanabha

The recent advances in the field of deep learning have not been fully utilised for decoding imagined speech primarily because of the unavailability of sufficient training samples to train a deep network. In this paper, we present a novel architecture that employs deep neural network (DNN) for classifying the words "in" and "cooperate" from the corresponding EEG signals in the ASU imagined speech dataset. Nine EEG channels, which best capture the underlying cortical activity, are chosen using common spatial pattern (CSP) and are treated as independent data vectors. Discrete wavelet transform (DWT) is used for feature extraction. To the best of our knowledge, so far DNN has not been employed as a classifier in decoding imagined speech. Treating the selected EEG channels corresponding to each imagined word as independent data vectors helps in providing sufficient number of samples to train a DNN. For each test trial, the final class label is obtained by applying a majority voting on the classification results of the individual channels considered in the trial. We have achieved accuracies comparable to the state-of-the-art results. The results can be further improved by using a higher-density EEG acquisition system in conjunction with other deep learning techniques such as long short-term memory.

Jerrin Thomas Panachakel, A. G. Ramakrishnan, A. G. Ramakrishnan

This paper proposes a novel approach that uses deep neural networks for classifying imagined speech, significantly increasing the classification accuracy. The proposed approach employs only the EEG channels over specific areas of the brain for classification, and derives distinct feature vectors from each of those channels. This gives us more data to train a classifier, enabling us to use deep learning approaches. Wavelet and temporal domain features are extracted from each channel. The final class label of each test trial is obtained by applying a majority voting on the classification results of the individual channels considered in the trial. This approach is used for classifying all the 11 prompts in the KaraOne dataset of imagined speech. The proposed architecture and the approach of treating the data have resulted in an average classification accuracy of 57.15%, which is an improvement of around 35% over the state-of-the-art results.

A Madhavaraj, T V Ananthapadmanabha, A G Ramakrishnan

Subjective and objective experiments are conducted to understand the extent to which a speaker's gender influences the acoustics of unvoiced (U) sounds. U segments of utterances are replaced by the corresponding segments of a speaker of opposite gender to prepare modified utterances. Humans are asked to judge if the modified utterance is spoken by one or two speakers. The experiments show that human subjects are unable to distinguish the modified from the original. Thus, listeners are able to identify the U segments irrespective of the gender, which may be based on some speaker-independent invariant acoustic cues. To test if this finding is purely a perceptual phenomenon, objective experiments are also conducted. Gender specific HMM based phoneme recognition systems are trained using the TIMIT training set and tested on (a) utterances spoken by the same gender (b) utterances spoken by the opposite gender and (c) the modified utterances of the test set. As expected, the performance is the highest for case (a) and the lowest for case (b). The performance degrades only slightly for case (c). This result shows that the speaker's gender does not as strongly influence the acoustics of U sounds as they do the voiced sounds.

K V Vijay Girish, A G Ramakrishnan, T V Ananthapadmanabha

A judicious combination of dictionary learning methods, block sparsity and source recovery algorithm are used in a hierarchical manner to identify the noises and the speakers from a noisy conversation between two people. Conversations are simulated using speech from two speakers, each with a different background noise, with varied SNR values, down to -10 dB. Ten each of randomly chosen male and female speakers from the TIMIT database and all the noise sources from the NOISEX database are used for the simulations. For speaker identification, the relative value of weights recovered is used to select an appropriately small subset of the test data, assumed to contain speech. This novel choice of using varied amounts of test data results in an improvement in the speaker recognition rate of around 15% at SNR of 0 dB. Speech and noise are separated using dictionaries of the estimated speaker and noise, and an improvement of signal to distortion ratios of up to 10% is achieved at SNR of 0 dB. K-medoid and cosine similarity based dictionary learning methods lead to better recognition of the background noise and the speaker. Experiments are also conducted on cases, where either the background noise or the speaker is outside the set of trained dictionaries. In such cases, adaptive dictionary learning leads to performance comparable to the other case of complete dictionaries.

T. V. Ananthapadmanabha, A. G. Ramakrishnan

Using a known speaker-intrinsic normalization procedure, formant data are scaled by the reciprocal of the geometric mean of the first three formant frequencies. This reduces the influence of the talker but results in a distorted vowel space. The proposed speaker-extrinsic procedure re-scales the normalized values by the mean formant values of vowels. When tested on the formant data of vowels published by Peterson and Barney, the combined approach leads to well separated clusters by reducing the spread due to talkers. The proposed procedure performs better than two top-ranked normalization procedures based on the accuracy of vowel classification as the objective measure.

K V Vijay Girish, T V Ananthapadmanabha, A G Ramakrishnan

A dictionary learning based audio source classification algorithm is proposed to classify a sample audio signal as one amongst a finite set of different audio sources. Cosine similarity measure is used to select the atoms during dictionary learning. Based on three objective measures proposed, namely, signal to distortion ratio (SDR), the number of non-zero weights and the sum of weights, a frame-wise source classification accuracy of 98.2% is obtained for twelve different sources. Cent percent accuracy has been obtained using moving SDR accumulated over six successive frames for ten of the audio sources tested, while the two other sources require accumulation of 10 and 14 frames.

T. V. Ananthapadmanabha, A. G. Ramakrishnan, Shubham Sharma

An objective critical distance (OCD) has been defined as that spacing between adjacent formants, when the level of the valley between them reaches the mean spectral level. The measured OCD lies in the same range (viz., 3-3.5 bark) as the critical distance determined by subjective experiments for similar experimental conditions. The level of spectral valley serves a purpose similar to that of the spacing between the formants with an added advantage that it can be measured from the spectral envelope without an explicit knowledge of formant frequencies. Based on the relative spacing of formant frequencies, the level of the spectral valley, VI (between F1 and F2) is much higher than the level of VII (spectral valley between F2 and F3) for back vowels and vice-versa for front vowels. Classification of vowels into front/back distinction with the difference (VI-VII) as an acoustic feature, tested using TIMIT, NTIMIT, Tamil and Kannada language databases gives, on the average, an accuracy of about 95%, which is comparable to the accuracy (90.6%) obtained using a neural network classifier trained and tested using MFCC as the feature vector for TIMIT database. The acoustic feature (VI-VII) has also been tested for its robustness on the TIMIT database for additive white and babble noise and an accuracy of about 95% has been obtained for SNRs down to 25 dB for both types of noise.

T. V. Ananthapadmanabha, A. G. Ramakrishnan, Pradeep Balachandran

Linear prediction (LP) technique estimates an optimum all-pole filter of a given order for a frame of speech signal. The coefficients of the all-pole filter, 1/A(z) are referred to as LP coefficients (LPCs). The gain of the inverse of the all-pole filter, A(z) at z = 1, i.e, at frequency = 0, A(1) corresponds to the sum of LPCs, which has the property of being lower (higher) than a threshold for the sonorants (fricatives). When the inverse-tan of A(1), denoted as T(1), is used a feature and tested on the sonorant and fricative frames of the entire TIMIT database, an accuracy of 99.07% is obtained. Hence, we refer to T(1) as sonorant-fricative discrimination index (SFDI). This property has also been tested for its robustness for additive white noise and on the telephone quality speech of the NTIMIT database. These results are comparable to, or in some respects, better than the state-of-the-art methods proposed for a similar task. Such a property may be used for segmenting a speech signal or for non-uniform frame-rate analysis.

T V Ananthapadmanabha, K V Vijay Girish, A G Ramakrishnan

Detection of transitions between broad phonetic classes in a speech signal is an important problem which has applications such as landmark detection and segmentation. The proposed hierarchical method detects silence to non-silence transitions, high amplitude (mostly sonorants) to low ampli- tude (mostly fricatives/affricates/stop bursts) transitions and vice-versa. A subset of the extremum (minimum or maximum) samples between every pair of successive zero-crossings is selected above a second pass threshold, from each bandpass filtered speech signal frame. Relative to the mid-point (reference) of a frame, locations of the first and the last extrema lie on either side, if the speech signal belongs to a homogeneous segment; else, both these locations lie on the left or the right side of the reference, indicating a transition frame. When tested on the entire TIMIT database, of the transitions detected, 93.6% are within a tolerance of 20 ms from the hand labeled boundaries. Sonorant, unvoiced non-sonorant and silence classes and their respective onsets are detected with an accuracy of about 83.5% for the same tolerance. The results are as good as, and in some respects better than the state-of-the-art methods for similar tasks.