Deep Neural Network (DNN) is a powerful machine learning model that has been successfully applied to a wide range of pattern classification tasks. Due to the great ability of the DNNs in learning complex mapping functions, it has been possible to train and deploy DNNs pretty much as a black box without the need to have an in-depth understanding of the inner workings of the model. However, this often leads to solutions and systems that achieve great performance, but offer very little in terms of how and why they work. This paper introduces Sensitivity-characterised Activity Neorogram (SCAN), a novel approach for understanding the inner workings of a DNN by analysing and visualising the sensitivity patterns of the neuron activities. SCAN constructs a low-dimensional visualisation space for the neurons so that the neuron activities can be visualised in a meaningful and interpretable way. The embedding of the neurons within this visualisation space can be used to compare the neurons, both within the same DNN and across different DNNs trained for the same task. This paper will present the observations from using SCAN to analyse DNN acoustic models for automatic speech recognition.
Among various training concepts for speaker adaptation, Speaker Adaptive Training (SAT) has been successfully applied to a standard Hidden Markov Model (HMM) speech recognizer, whose state is associated with Gaussian Mixture Models (GMMs). On the other hand, focusing on the high discriminative power of Deep Neural Networks (DNNs), a new type of speech recognizer structure, which combines DNNs and HMMs, has been vigorously investigated in the speaker adaptation research field. Along these two lines, it is natural to conceive of further improvement to a DNN-HMM recognizer by employing the training concept of SAT. In this paper, we propose a novel speaker adaptation scheme that applies SAT to a DNN-HMM recognizer. Our SAT scheme allocates a Speaker Dependent (SD) module to one of the intermediate layers of DNN, treats its remaining layers as a Speaker Independent (SI) module, and jointly trains the SD and SI modules while switching the SD module in a speaker-by-speaker manner. We implement the scheme using a DNN-HMM recognizer, whose DNN has seven layers, and elaborate its utility over TED Talks corpus data. Our experimental results show that in the supervised adaptation scenario, our Speaker-Adapted (SA) SAT-based recognizer reduces the word error rate of the baseline SI recognizer and the lowest word error rate of the SA SI recognizer by 8.4% and 0.7%, respectively, and by 6.4% and 0.6% in the unsupervised adaptation scenario. The error reductions gained by our SA-SAT-based recognizers proved to be significant by statistical testing. The results also show that our SAT-based adaptation outperforms, regardless of the SD module layer selection, its counterpart SI-based adaptation, and that the inner layers of DNN seem more suitable for SD module allocation than the outer layers.
Audio-Visual Speech Recognition (AVSR) is one of techniques to enhance robustness of speech recognizer in noisy or real environments. On the other hand, Deep Neural Networks (DNNs) have recently attracted a lot of attentions of researchers in the speech recognition field, because we can drastically improve recognition performance by using DNNs. There are two ways to employ DNN techniques for speech recognition: a hybrid approach and a tandem approach; in the hybrid approach an emission probability on each Hidden Markov Model (HMM) state is computed using a DNN, while in the tandem approach a DNN is composed into a feature extraction scheme. In this paper, we investigate and compare several DNN-based AVSR methods to mainly clarify how we should incorporate audio and visual modalities using DNNs. We carried out recognition experiments using a corpus CENSREC-1-AV, and we discuss the results to find out the best DNN-based AVSR modeling. Then it turns out that a tandem-based method using audio Deep Bottle-Neck Features (DBNFs) and visual ones with multi-stream HMMs is the most suitable, followed by a hybrid approach and another tandem scheme using audio-visual DBNFs.
This paper aims to investigate the performance improvements made possible by combining various major language model (LM) technologies together and to reveal the interactions between LM technologies in spontaneous automatic speech recognition tasks. While it is clear that recent practical LMs have several problems, isolated use of major LM technologies does not appear to offer sufficient performance. In consideration of this fact, combining various LM technologies has been also examined. However, previous works only focused on modeling technologies with limited text resources, and did not consider other important technologies in practical language modeling, i.e., use of external text resources and unsupervised adaptation. This paper, therefore, employs not only manual transcriptions of target speech recognition tasks but also external text resources. In addition, unsupervised LM adaptation based on multi-pass decoding is also added to the combination. We divide LM technologies into three categories and employ key ones including recurrent neural network LMs or discriminative LMs. Our experiments show the effectiveness of combining various LM technologies in not only in-domain tasks, the subject of our previous work, but also out-of-domain tasks. Furthermore, we also reveal the relationships between the technologies in both tasks.
This paper aims to improve the domain robustness of language modeling for automatic speech recognition (ASR). To this end, we focus on applying the latent words language model (LWLM) to ASR. LWLMs are generative models whose structure is based on Bayesian soft class-based modeling with vast latent variable space. Their flexible attributes help us to efficiently realize the effects of smoothing and dimensionality reduction and so address the data sparseness problem; LWLMs constructed from limited domain data are expected to robustly cover unknown multiple domains in ASR. However, the attribute flexibility seriously increases computation complexity. If we rigorously compute the generative probability for an observed word sequence, we must consider the huge quantities of all possible latent word assignments. Since this is computationally impractical, some approximation is inevitable for ASR implementation. To solve the problem and apply this approach to ASR, this paper presents an n-gram approximation of LWLM. The n-gram approximation is a method that approximates LWLM as a simple back-off n-gram structure, and offers LWLM-based robust one-pass ASR decoding. Our experiments verify the effectiveness of our approach by evaluating perplexity and ASR performance in not only in-domain data sets but also out-of-domain data sets.
Building high-quality text-to-speech (TTS) systems without expert knowledge of the target language and/or time-consuming manual annotation of speech and text data is an important yet challenging research topic. In this kind of TTS system, it is vital to find representation of the input text that is both effective and easy to acquire. Recently, the continuous representation of raw word inputs, called “word embedding”, has been successfully used in various natural language processing tasks. It has also been used as the additional or alternative linguistic input features to a neural-network-based acoustic model for TTS systems. In this paper, we further investigate the use of this embedding technique to represent phonemes, syllables and phrases for the acoustic model based on the recurrent and feed-forward neural network. Results of the experiments show that most of these continuous representations cannot significantly improve the system's performance when they are fed into the acoustic model either as additional component or as a replacement of the conventional prosodic context. However, subjective evaluation shows that the continuous representation of phrases can achieve significant improvement when it is combined with the prosodic context as input to the acoustic model based on the feed-forward neural network.
This paper describes a novel statistical bandwidth extension (BWE) technique based on a Gaussian mixture model (GMM) and a sub-band basis spectrum model (SBM), in which each dimensional component represents a specific acoustic space in the frequency domain. The proposed method can achieve the BWE from speech data with an arbitrary frequency bandwidth whereas the conventional methods perform the conversion from fixed narrow-band data. In the proposed method, we train a GMM with SBM parameters extracted from full-band spectra in advance. According to the bandwidth of input signal, the trained GMM is reconstructed to the GMM of the joint probability density between low-band SBM and high-band SBM components. Then high-band SBM components are estimated from low-band SBM components of the input signal based on the reconstructed GMM. Finally, BWE is achieved by adding the spectra decoded from estimated high-band SBM components to the ones of the input signal. To construct the full-band signal from the narrow-band one, we apply this method to log-amplitude spectra and aperiodic components. Objective and subjective evaluation results show that the proposed method extends the bandwidth of speech data robustly for the log-amplitude spectra. Experimental results also indicate that the aperiodic component extracted from the upsampled narrow-band signal realizes the same performance as the restored and the full-band aperiodic components in the proposed method.
This paper presents a novel statistical sample-based approach for Gaussian Mixture Model (GMM)-based Voice Conversion (VC). Although GMM-based VC has the promising flexibility of model adaptation, quality in converted speech is significantly worse than that of natural speech. This paper addresses the problem of inaccurate modeling, which is one of the main reasons causing the quality degradation. Recently, we have proposed statistical sample-based speech synthesis using rich context models for high-quality and flexible Hidden Markov Model (HMM)-based Text-To-Speech (TTS) synthesis. This method makes it possible not only to produce high-quality speech by introducing ideas from unit selection synthesis, but also to preserve flexibility of the original HMM-based TTS. In this paper, we apply this idea to GMM-based VC. The rich context models are first trained for individual joint speech feature vectors, and then we gather them mixture by mixture to form a Rich context-GMM (R-GMM). In conversion, an iterative generation algorithm using R-GMMs is used to convert speech parameters, after initialization using over-trained probability distributions. Because the proposed method utilizes individual speech features, and its formulation is the same as that of conventional GMM-based VC, it makes it possible to produce high-quality speech while keeping flexibility of the original GMM-based VC. The experimental results demonstrate that the proposed method yields significant improvements in term of speech quality and speaker individuality in converted speech.
The optimization of spoken dialog management policies is a non-trivial task due to the erroneous inputs from speech recognition and language understanding modules. The dialog manager needs to ground uncertain semantic information at times to fully understand the need of human users and successfully complete the required dialog tasks. Approaches based on reinforcement learning are currently mainstream in academia and have been proved to be effective, especially when operating in noisy environments. However, in reinforcement learning the dialog strategy is often represented by complex numeric model and thus is incomprehensible to humans. The trained policies are very difficult for dialog system designers to verify or modify, which largely limits the deployment for commercial applications. In this paper we propose a novel framework for optimizing dialog policies specified in human-readable domain language using genetic algorithm. We present learning algorithms using user simulator and real human-machine dialog corpora. Empirical experimental results show that the proposed approach can achieve competitive performance on par with some state-of-the-art reinforcement learning algorithms, while maintaining a comprehensible policy structure.
In this work, we propose a new statistical model for building robust dialog systems using neural networks to either retrieve or generate dialog response based on an existing data sources. In the retrieval task, we propose an approach that uses paraphrase identification during the retrieval process. This is done by employing recursive autoencoders and dynamic pooling to determine whether two sentences with arbitrary length have the same meaning. For both the generation and retrieval tasks, we propose a model using long short term memory (LSTM) neural networks that works by first using an LSTM encoder to read in the user's utterance into a continuous vector-space representation, then using an LSTM decoder to generate the most probable word sequence. An evaluation based on objective and subjective metrics shows that the new proposed approaches have the ability to deal with user inputs that are not well covered in the database compared to standard example-based dialog baselines.
This study proposes a two-pass spoken term detection (STD) method. The first pass uses a phoneme-based dynamic time warping (DTW)-based STD, and the second pass recomputes detection scores produced by the first pass using conditional random fields (CRF)-based triphone detectors. In the second-pass, we treat STD as a sequence labeling problem. We use CRF-based triphone detection models based on features generated from multiple types of phoneme-based transcriptions. The models train recognition error patterns such as phoneme-to-phoneme confusions in the CRF framework. Consequently, the models can detect a triphone comprising a query term with a detection probability. In the experimental evaluation of two types of test collections, the CRF-based approach worked well in the re-ranking process for the DTW-based detections. CRF-based re-ranking showed 2.1% and 2.0% absolute improvements in F-measure for each of the two test collections.
This study presents a two-stage spoken term detection (STD) method that uses the same STD engine twice and a support vector machine (SVM)-based classifier to verify detected terms from the STD engine's output. In a front-end process, the STD engine is used to pre-index target spoken documents from a keyword list built from an automatic speech recognition result. The STD result includes a set of keywords and their detection intervals (positions) in the spoken documents. For keywords having competitive intervals, we rank them based on the STD matching cost and select the one having the longest duration among competitive detections. The selected keywords are registered in the pre-index. They are then used to train an SVM-based classifier. In a query term search process, a query term is searched by the same STD engine, and the output candidates are verified by the SVM-based classifier. Our proposed two-stage STD method with pre-indexing was evaluated using the NTCIR-10 SpokenDoc-2 STD task and it drastically outperformed the traditional STD method based on dynamic time warping and a confusion network-based index.
We propose a novel method for estimating acoustic scenes such as user activities, e.g., “cooking,” “vacuuming,” “watching TV,” or situations, e.g., “being on the bus,” “being in a park,” “meeting,” utilizing the information of acoustic events. There are some methods for estimating acoustic scenes that associate a combination of acoustic events with an acoustic scene. However, the existing methods cannot adequately express acoustic scenes, e.g., “cooking,” that have more than one subordinate category, e.g., “frying ingredients” or “plating food,” because they directly associate acoustic events with acoustic scenes. In this paper, we propose an acoustic scene estimation method based on a hierarchical probabilistic generative model of an acoustic event sequence taking into account the relation among acoustic scenes, their subordinate categories, and acoustic event sequences. In the proposed model, each acoustic scene is represented as a probability distribution over their unsupervised subordinate categories, called “acoustic sub-topics,” and each acoustic sub-topic is represented as a probability distribution over acoustic events. Acoustic scene estimation experiments with real-life sounds showed that the proposed method could correctly extract subordinate categories of acoustic scenes.
The introduction of deep neural networks (DNNs) leads to a significant improvement of the automatic speech recognition (ASR) performance. However, the whole ASR system remains sophisticated due to the dependent on the hidden Markov model (HMM). Recently, a new end-to-end ASR framework, which utilizes recurrent neural networks (RNNs) to directly model context-independent targets with connectionist temporal classification (CTC) objective function, is proposed and achieves comparable results with the hybrid HMM/DNN system. In this paper, we investigate per-dimensional learning rate methods, ADAGRAD and ADADELTA included, to improve the recognition of the end-to-end system, based on the fact that the blank symbol used in CTC technique dominates the output and these methods give frequent features small learning rates. Experiment results show that more than 4% relative reduction of word error rate (WER) as well as 5% absolute improvement of label accuracy on the training set are achieved when using ADADELTA, and fewer epochs of training are needed.
Multi-task learning in deep neural networks has been proven to be effective for acoustic modeling in speech recognition. In the paper, this technique is applied to Mandarin-English code-mixing recognition. For the primary task of the senone classification, three schemes of the auxiliary tasks are proposed to introduce the language information to networks and improve the prediction of language switching. On the real-world Mandarin-English test corpus in mobile voice search, the proposed schemes enhanced the recognition on both languages and reduced the relative overall error rates by 3.5%, 3.8% and 5.8% respectively.
This paper proposes a new quantization framework on activation function of deep neural networks (DNN). We implement fixed-point DNN by quantizing the activations into powers-of-two integers. The costly multiplication operations in using DNN can be replaced with low-cost bit-shifts to massively save computations. Thus, applying DNN-based speech recognition on embedded systems becomes much easier. Experiments show that the proposed method leads to no performance degradation.
Short texts usually encounter the problem of data sparseness, as they do not provide sufficient term co-occurrence information. In this paper, we show how to mitigate the problem in short text classification through word embeddings. We assume that a short text document is a specific sample of one distribution in a Gaussian-Bayesian framework. Furthermore, a fast clustering algorithm is utilized to expand and enrich the context of short text in embedding space. This approach is compared with those based on the classical bag-of-words approaches and neural network based methods. Experimental results validate the effectiveness of the proposed method.
We consider the problem of sparse signal recovery from 1-bit measurements. Due to the noise present in the acquisition and transmission process, some quantized bits may be flipped to their opposite states. These sign flips may result in severe performance degradation. In this study, a novel algorithm, termed HISTORY, is proposed. It consists of Hamming support detection and coefficients recovery. The HISTORY algorithm has high recovery accuracy and is robust to strong measurement noise. Numerical results are provided to demonstrate the effectiveness and superiority of the proposed algorithm.
Given a graph G=(V,E) where V and E are a vertex and an edge set, respectively, specified with a subset VNT of vertices called a non-terminal set, the spanning tree with non-terminal set VNT is a connected and acyclic spanning subgraph of G that contains all the vertices of V where each vertex in a non-terminal set is not a leaf. In the case where each edge has the weight of a nonnegative integer, the problem of finding a minimum spanning tree with a non-terminal set VNT of G was known to be NP-hard. However, the complexity of finding a spanning tree on general graphs where each edge has the weight of one was unknown. In this paper, we consider this problem and first show that it is NP-hard even if each edge has the weight of one on general graphs. We also show that if G is a cograph then finding a spanning tree with a non-terminal set VNT of G is linearly solvable when each edge has the weight of one.
This article presents efficient strategies for evacuating from an unknown affected area in a plane. Evacuation is the process of movement away from a threat or hazard such as natural disasters. Consider that one or n(n ≥ 3) agents are lost in an unknown convex region P. The agents know neither the boundary information of P nor their positions. We seek competitive strategies that can evacuate the agent from P as quickly as possible. The performance of the strategy is measured by a competitive ratio of the evacuation path over the shortest path. We give a 13.812-competitive spiral strategy for one agent, and prove that it is optimal among all monotone and periodic strategies by showing a matching lower bound. Also, we give a new competitive strategy EES for n(n ≥ 3) agents and adjust it to be more efficient with the analysis of its performance.
Embedded memory is extensively being used in SoCs, and is rapidly growing in size and density. It contributes to SoCs to have greater features, but at the expense of taking up the most area. Due to continuous scaling of nanoscale device technology, large area size memory introduces aging-induced faults and soft errors, which affects reliability. In-field test and repair, as well as ECC, can be used to maintain reliability, and recently, these methods are used together to form a combined approach, wherein uncorrectable words are repaired, while correctable words are left to the ECC. In this paper, we propose a novel in-field repair strategy that repairs uncorrectable words, and possibly correctable words, for an ECC-based memory architecture. It executes an adaptive reconfiguration method that ensures 'fresh' memory words are always used until spare words run out. Experimental results demonstrate that our strategy enhances reliability, and the area overhead contribution is small.
Collaborative filtering (CF) has been widely used in recommender systems to generate personalized recommendations. However, recommender systems using CF are vulnerable to shilling attacks, in which attackers inject fake profiles to manipulate recommendation results. Thus, shilling attacks pose a threat to the credibility of recommender systems. Previous studies mainly derive features from characteristics of item ratings in user profiles to detect attackers, but the methods suffer from low accuracy when attackers adopt new rating patterns. To overcome this drawback, we derive features from properties of item popularity in user profiles, which are determined by users' different selecting patterns. This feature extraction method is based on the prior knowledge that attackers select items to rate with man-made rules while normal users do this according to their inner preferences. Then, machine learning classification approaches are exploited to make use of these features to detect and remove attackers. Experiment results on the MovieLens dataset and Amazon review dataset show that our proposed method improves detection performance. In addition, the results justify the practical value of features derived from selecting patterns.
With the surge of social media platform, users' profile information become treasure to enhance social network services. However, attributes information of most users are not complete, thus it is important to infer latent attributes of users. Contemporary attribute inference methods have a basic assumption that there are enough labeled data to train a model. However, in social media, it is very expensive and difficult to label a large amount of data. In this paper, we study the latent attribute inference problem with very small labeled data and propose the SRW-COND solution. In order to solve the difficulty of small labeled data, SRW-COND firstly extends labeled data with a simple but effective greedy algorithm. Then SRW-COND employs a supervised random walk process to effectively utilize the known attributes information and link structure of users. Experiments on two real datasets illustrate the effectiveness of SRW-COND.
This paper presents a semi-incremental recognition method for on-line handwritten Japanese text and its evaluation. As text becomes longer, recognition time and waiting time become large if it is recognized after it is written (batch recognition). Thus, incremental methods have been proposed with recognition triggered by every stroke but the recognition rates are damaged and more CPU time is incurred. We propose semi-incremental recognition and employ a local processing strategy by focusing on a recent sequence of strokes defined as ”scope” rather than every new stroke. For the latest scope, we build and update a segmentation and recognition candidate lattice and advance the best-path search incrementally. We utilize the result of the best-path search in the previous scope to exclude unnecessary segmentation candidates. This reduces the number of candidate character recognition with the result of reduced processing time. We also reuse the segmentation and recognition candidate lattice in the previous scope for the latest scope. Moreover, triggering recognition processes every several strokes saves CPU time. Experiments made on TUAT-Kondate database show the effectiveness of the proposed semi-incremental recognition method not only in reduced processing time and waiting time, but also in recognition accuracy.
Estimating the cardinality of flows over sliding windows on high-speed links is still a challenging work under time and space constrains. To solve this problem, we present a novel data structure maintaining a summary of data and propose a constant-time update algorithm for fast evicting expired information. Moreover, a further memory-reducing schema is given at a cost of very little loss of accuracy.
NAND flash memory has been widely used in storage systems. Aiming to design an efficient buffer policy for NAND flash memory, a life-aware buffer management algorithm named LAB-LRU is proposed, which manages the buffer by three LRU lists. A life value is defined for every page and the active pages with higher life value can stay longer in the buffer. The definition of life value considers the effect of access frequency, recency and the cost of flash read and write operations. A series of trace-driven simulations are carried out and the experimental results show that the proposed LAB-LRU algorithm outperforms the previous best-known algorithms significantly in terms of the buffer hit ratio, the numbers of flash write and read operations and overall runtime.
Public auditing is a new technique to protect the integrity of outsourced data in the remote cloud. Users delegate the ability of auditing to a third party auditor (TPA), and assume that each result from the TPA is correct. However, the TPA is not always trustworthy in reality. In this paper, we consider a scenario in which the TPA may lower the reputation of the cloud server by cheating users, and propose a novel public auditing scheme to address this security issue. The analyses and the evaluation prove that our scheme is both secure and efficient.
Traditional action recognition approaches use pre-defined rigid areas to process the space-time information, e.g. spatial pyramids, cuboids. However, most action categories happen in an unconstrained manner, that is, the same action in different videos can happen at different places. Thus we need a better video representation to deal with the space-time variations. In this paper, we introduce the idea of mining spatial temporal saliency. To better handle the uniqueness of each video, we use a space-time over-segmentation approach, e.g. supervoxel. We choose three different saliency measures that take not only the appearance cues, but also the motion cues into consideration. Furthermore, we design a category-specific mining process to find the discriminative power in each action category. Experiments on action recognition datasets such as UCF11 and HMDB51 show that the proposed spatial temporal saliency video representation can match or surpass some of the state-of-the-art alternatives in the task of action recognition.
In practice, emotional speech utterances are often collected from different devices or conditions, which will lead to discrepancy between the training and testing data, resulting in sharp decrease of recognition rates. To solve this problem, in this letter, a novel transfer semi-supervised non-negative matrix factorization (TSNMF) method is presented. A semi-supervised negative matrix factorization algorithm, utilizing both labeled source and unlabeled target data, is adopted to learn common feature representations. Meanwhile, the maximum mean discrepancy (MMD) as a similarity measurement is employed to reduce the distance between the feature distributions of two databases. Finally, the TSNMF algorithm, which optimizes the SNMF and MMD functions together, is proposed to obtain robust feature representations across databases. Extensive experiments demonstrate that in comparison to the state-of-the-art approaches, our proposed method can significantly improve the cross-corpus recognition rates.
As an important extension of high-efficiency video coding (HEVC), screen content coding (SCC) includes various new coding modes, such as Intra Block Copy (IBC), Palette-based coding (Palette), and Adaptive Color Transform (ACT). These new tools have improved screen content encoding performance. This paper proposed a novel and fast algorithm by classifying Code Units (CUs) as text CUs or non-text CUs. For text CUs, the Intra mode was skipped in the compression process, whereas for non-text CUs, the IBC mode was skipped. The current CU depth range was then predicted according to its adjacent left CU depth level. Compared with the reference software HM16.7+SCM5.4, the proposed algorithm reduced encoding time by 23% on average and achieved an approximate 0.44% increase in Bjøntegaard delta bit rate and a negligible peak signal-to-noise ratio loss.
Encoding multiple SIFT descriptors into a single vector is a key technique for efficient object image retrieval. In this paper, we propose an extension of local coordinate system (LCS) for image representation. The previous LCS approaches encode each SIFT descriptor by a single local coordinate, which is not adequate for localizing its position in the descriptor space. Instead, we use multiple local coordinates to represent each descriptor with PCA-based decorrelation. Experiments show that this simple modification can improve retrieval performance significantly.
A novel method for illumination-invariant face representation is presented based on the orthogonal decomposition of the local image structure. One important advantage of the proposed method is that image gradients and corresponding intensity values are simultaneously used with our decomposition procedure to preserve the original texture while yielding the illumination-invariant feature space. Experimental results demonstrate that the proposed method is effective for face recognition and verification even with diverse lighting conditions.
We propose a new visual tracking method, where the target appearance is represented by combining color distribution and keypoints. Firstly, the object is localized via a keypoint-based tracking and matching strategy, where a new clustering method is presented to remove outliers. Secondly, the tracking confidence is evaluated by the color template. According to the tracking confidence, the local and global keypoints matching can be performed adaptively. Finally, we propose a target appearance update method in which the new appearance can be learned and added to the target model. The proposed tracker is compared with five state-of-the-art tracking methods on a recent benchmark dataset. Both qualitative and quantitative evaluations show that our method has favorable performance.
Automated biometrics identification using finger vein images has increasingly generated interest among researchers with emerging applications in human biometrics. The traditional feature-level fusion strategy is limited and expensive. To solve the problem, this paper investigates the possible use of infrared hybrid finger patterns on the back side of a finger, which includes both the information of finger vein and finger dorsal textures in original image, and a database using the proposed hybrid pattern is established. Accordingly, an Intersection enhanced Gabor based Direction Coding (IGDC) method is proposed. The Experiment achieves a recognition ratio of 98.4127% and an equal error rate of 0.00819 on our newly established database, which is fairly competitive.