Due to the benefits of palmprint recognition and the advantages of biometric fusion systems, it is necessary to study multi-source palmprint fusion systems. Unfortunately, the research on multi-instance palmprint feature fusion is absent until now. In this paper, we extract the features of left and right palmprints with two-dimensional discrete cosine transform (2DDCT) to constitute a dual-source space. Normalization is utilized in dual-source space to avoid the disturbance caused by the coefficients with large absolute values. Thus complicated pre-masking is needless and arbitrary removing of discriminative coefficients is avoided. Since more discriminative coefficients can be preserved and retrieved with discrimination power analysis (DPA) from dual-source space, the accuracy performance is improved. The experiments performed on contactless palmprint database confirm that dual-source DPA, which is designed for multi-instance palmprint feature fusion recognition, outperforms single-source DPA.

Dual-source discrimination power analysis for multi-instance contactless palmprint recognition

Although Discrete Cosine Transform (DCT) is widely employed to extract proper features for biometric recognition, the problem on how to select proper DCT coefficients to obtain the best discrimination effect has not been solved satisfactorily. Some approaches discard the low-frequency DCT coefficients unreasonably and rely on proper premasking window to improve performance. But there is not a uniform criterion to optimize the shape and size of the premasking window, so it is an inconvenient processing for coefficient selection.Three processes, used to enhance discriminant ability in DCT domain, and the relationship between them are summarized and discussed systematically. Furthermore, this paperexplains the phenomenon why the recognition rate is low without discarding the low-frequency DCT coefficients reasonably and then proposes dynamic weighted discrimination power analysis (DWDPA) to enhance the discrimination power (DP) of the selected DCT coefficients. DWDPA does not need premasking window and preserves more DCT coefficients with higher DP. Normalization prevents the DCT coefficients with large absolute values from destroying the DP of the other DCT coefficients that have less absolute values but high DP values. The DCT coefficients with larger DP values are given larger weights adaptively to optimize and enhance the recognition performance. The experiments on ORL, Yale and PolyU databases captured by biometric sensors prove the advantages of DWDPA obviously.

	 

	Key words: Dynamic weighted discrimination power analysis (DWDPA), discrete cosine transforms (DCT), biometric sensors, face recognition, palmprint recognition.

Dynamic weighted discrimination power analysis: A novel approach for face and palmprint recognition in DCT domain

Speech recognition in noisy environments

Two popular representation learning paradigms are dictionary learning and deep learning. While dictionary learning focuses on learning “basis” and “features” by matrix factorization, deep learning focuses on extracting features via learning “weights” or “filter” in a greedy layer by layer fashion. This paper focuses on combining the concepts of these two paradigms by proposing deep dictionary learning and show how deeper architectures can be built using the layers of dictionary learning. The proposed technique is compared with other deep learning approaches, such as stacked autoencoder, deep belief network, and convolutional neural network. Experiments on benchmark data sets show that the proposed technique achieves higher classification and clustering accuracies. On a real-world problem of electrical appliance classification, we show that deep dictionary learning excels where others do not yield at-par performance. We postulate that the proposed formulation can pave the path for a new class of deep learning tools.

Deep dictionary learning

Discrimination power analysis (DPA) is a statistical analysis combining discrimination concept with discrete cosine transform coefficients (DCTCs) properties Unfortunately there is not a uniform and effective criterion to optimize the shape and size of premasking window on which the effect of DPA excessively relies Proper premasking is an auxiliary process to select the feature coefficients that have more discrimination power (DP) Dynamic weighted DPA (DWDPA) is proposed in this paper to enhance the DP of the selected DCTCs without premasking window, in other words, it does not need to optimize the shape and size of premasking window The DCTCs are adaptively selected according to their discrimination power values (DPVs) More DCTCs with higher DP are preserved The selected coefficients are normalized and dynamic weighted according to their DPVs Normalization assures that the DCTCs with large absolute value don't destroy the DP of the other DCTCs that have less absolute value but high DPVs Dynamic weighting gives larger weights to the DCTCs with larger DPVs which optimizes and enhances the recognition performance The experimental results on ORL, Yale and PolyU databases show that DWDPA outperforms DPA obviously

Dynamic weighted discrimination power analysis in DCT domain for face and palmprint recognition

Feature extraction using discrete cosine transform and discrimination power analysis with a face recognition technology

In this paper, a meter classification system has been proposed for Persian poems based on features extracted from uttered poem. In the first stage, the utterance has been segmented into syllables using three features, pitch frequency and modified energy of each frame of the utterance and its temporal variations. In the second stage, each syllable is classified into long syllable and short syllable classes which is a historically convenient categorization in Persian literature. In this stage, the classifier is an SVM classifier with radial basis function kernel and employed features are the syllable temporal duration, zero crossing rate and PARCOR coefficients of each syllable. The sequence of extracted syllables classes is then compared with classic Persian meter styles using dynamic time warping, to make the system robust against syllables insertion, deletion or classification. The system has been evaluated on 136 poetries utterances from 12 Persian meter styles gathered from 8 speakers, using k-fold evaluation strategy. The results show 91% accuracy in three top meter style choices of the system.

Automatic meter classification in Persian poetries using support vector machines

This paper focuses on enhancing MFCC features using a set of MLP NN in order to improve phoneme recognition accuracy under different noise types and SNRs. A NN can be used in different domains (between any pair of MFCC feature extraction blocks). It includes FFT, MEL, LOG, DCT and DELTA domains. Various domains have different complexities and achieve different degrees. A comparative study is done in this paper in order to find the best domain. Furthermore, a set of MLP NNs, instead of one NN, is used to enhance various noise types with different levels of SNRs. In this case, each NN is trained with a special noise type and SNR. The database used in the simulations is created by artificially adding different types of noises from the NOISEX-92 database to a subset of TIMIT speech corpus. Our experiments show that the highest improvement is achievable in LOG domain. It is also shown that although the performance decreases slightly in the DCT domain, the complexity is reduced to one fourth in this domain.

Robust phoneme recognition using MLP neural networks in various domains of MFCC features

In this paper, we have proposed an efficient and effective nonlinear feature domain noise suppression algorithm, motivated by the minimum mean square error (MMSE) optimization criterion. A Multi Layer Perceptron (MLP) neural network in the log spectral domain has been employed to minimize the difference between noisy and clean speech. By using this method, as a pre-processing stage of a speech recognition system, the recognition rate in noisy environments has been improved. We extended the application of the system to different environments with different noises without retraining HMM model. We trained the feature extraction stage with a small portion of noisy data which was created by artificially adding different types of noises from the NOISEX-92 database to the TIMIT speech database. In real environment, where our speech recognition systems must work, different types of noises with various SNRs exist. Our proposed method suggests four strategies based on the system capability to identify the noise type and SNR. Experimental results show that the proposed method achieves significant improvement in recognition rates.

Robust speech recognition using MLP neural network in log-spectral domain

We propose an efficient and effective nonlinear feature domain noise suppression algorithm, motivated by the minimum mean square error (MMSE) optimization criterion. Multi Layer Perceptron (MLP) neural network in the log spectral domain minimizes the difference between noisy and clean speech. By using this method as a pre-processing stage of a speech recognition system, the recognition rate in noisy environments is improved. We can extend the application of the system to different environments with different noises without re-training it. We need only to train the preprocessing stage with a small portion of noisy data which is created by artificially adding different types of noises from the NOISEX-92 database to the TIMIT speech database. Experimental results show that the proposed method can achieve significant improvement of recognition rates.

Masoumeh P. Ghaemmaghami

Papers

Feature extraction using discrete cosine transform and discrimination power analysis with a face recognition technology

Automatic meter classification in Persian poetries using support vector machines

Robust phoneme recognition using MLP neural networks in various domains of MFCC features

Robust speech recognition using MLP neural network in log-spectral domain

Noise reduction algorithm for robust speech recognition using MLP neural network