Speaker-dependent Recognition Research Articles

Isolated word recognition based on PNCC with different classifiers in a noisy environment

Speech Recognition is one of the important methods to control robotics, mobile applications, automobile, etc. Nevertheless, still, the Speech recognition rate needs to be improved especially in noisy environments. The main goal of this paper is to investigate the best features for isolated word recognition between the most common features extraction technique Mel Frequency Cepstral Coefficient MFCC and noise immunity features extraction technique Power Normalized Cepstral Coefficient PNCC, with the aid of four classifiers, Weighted KNN, Medium KNN, Linear SVM, Quadratic SVM in different levels of noise. The database that is used for this study is Speech Commands Data Set v0.01 which is containing 64,727 audio files to get a good estimation of the recognition rate. The algorithm based on MFCC is recommended for Speaker dependent speech recognition a little bit better (1.1%) than the algorithm based on PNCC in a noise-free environment. Nevertheless, the proposed algorithm based on PNCC is more efficient than the algorithm based on MFCC for Speaker independent speech recognition, and overall Recognition accuracy. Moreover, it has more immunity to noise and it is more accurate than MFCC by 8.7% on average with already trained speakers and 18.528% on average with new trained speakers’ tests. Finally, the proposed algorithm based on the PNCC feature extraction with aid of Weighted KNN is more accurate than other algorithms for Speaker independent speech recognition and overall Recognition accuracy. Moreover, it has more immunity to noise.

Silent speech command word recognition using stepped frequency continuous wave radar

Recovering speech in the absence of the acoustic speech signal itself, i.e., silent speech, holds great potential for restoring or enhancing oral communication in those who lost it. Radar is a relatively unexplored silent speech sensing modality, even though it has the advantage of being fully non-invasive. We therefore built a custom stepped frequency continuous wave radar hardware to measure the changes in the transmission spectra during speech between three antennas, located on both cheeks and the chin with a measurement update rate of 100 Hz. We then recorded a command word corpus of 40 phonetically balanced, two-syllable German words and the German digits zero to nine for two individual speakers and evaluated both the speaker-dependent multi-session and inter-session recognition accuracies on this 50-word corpus using a bidirectional long-short term memory network. We obtained recognition accuracies of 99.17% and 88.87% for the speaker-dependent multi-session and inter-session accuracy, respectively. These results show that the transmission spectra are very well suited to discriminate individual words from one another, even across different sessions, which is one of the key challenges for fully non-invasive silent speech interfaces.

Effect on speech emotion classification of a feature selection approach using a convolutional neural network.

Speech emotion recognition (SER) is a challenging issue because it is not clear which features are effective for classification. Emotionally related features are always extracted from speech signals for emotional classification. Handcrafted features are mainly used for emotional identification from audio signals. However, these features are not sufficient to correctly identify the emotional state of the speaker. The advantages of a deep convolutional neural network (DCNN) are investigated in the proposed work. A pretrained framework is used to extract the features from speech emotion databases. In this work, we adopt the feature selection (FS) approach to find the discriminative and most important features for SER. Many algorithms are used for the emotion classification problem. We use the random forest (RF), decision tree (DT), support vector machine (SVM), multilayer perceptron classifier (MLP), and k-nearest neighbors (KNN) to classify seven emotions. All experiments are performed by utilizing four different publicly accessible databases. Our method obtains accuracies of 92.02%, 88.77%, 93.61%, and 77.23% for Emo-DB, SAVEE, RAVDESS, and IEMOCAP, respectively, for speaker-dependent (SD) recognition with the feature selection method. Furthermore, compared to current handcrafted feature-based SER methods, the proposed method shows the best results for speaker-independent SER. For EMO-DB, all classifiers attain an accuracy of more than 80% with or without the feature selection technique.

Effects of the Dynamic and Energy Based Feature Extraction on Hindi Speech Recognition

Background: Speech Recognition is the most effective and suitable way of communication. Extracted features play an important role in speech recognition. Previous research works for Hindi speech recognition lack detailed comparative analysis of the feature extraction methods using dynamic and energy parameters. Objective: The research work presents experimental work done to explore the effects of integrating dynamic coefficients and energy parameters with different feature extraction techniques on Connected word Hindi Speech recognition. As extracted features play a significant role in speech recognition, a comparative analysis is presented to show the effects of integration of dynamic and energy parameters to basic extracted features. Method: Speaker dependent system was proposed with monophones based five states Hidden Markov Model (HMM) using HTK Tool kit. Speech data set of connected words in Hindi was created. The feature extraction techniques such as Linear Predictive Coding Cepstral coefficients (LPCCs), Mel Frequency Cepstral Coefficients (MFCCs), and Perceptual Linear Prediction (PLPs) coefficients were applied integrating delta, delta2, and energy parameters to evaluate the performance of the proposed methodology for speaker dependent recognition. Results: Experimental results show that the system achieved the highest recognition word accuracy of 89.97% using PLP coefficients. The PLP coefficients achieved 4% increment in word accuracy than original MFCCs and a 16% increment in word accuracy than LPCCs. Adding energy parameters to original MFCCs increased word accuracy by 1.5% only while adding dynamic coefficients delta and double delta had no significant effect on speech recognition accuracy. Conclusion: Research findings reveal that PLP coefficients outperformed. Explorations reveal that the integration of energy parameters are better than original MFCCs. Investgations also reveal that adding energy parametres improved recognition score while adding delta and delta2 coefficients to basic features did not improve the recognition scores. Research findings could be used to enhance the performance of a speech recognition system by using a suitable feature extraction technique and combining the different feature extraction techniques. Further, investigations can be used to develop language resources for refining speech recognition. The work can be extended to develop a continuous Hindi speech recognition system.

A Proposed Speaker Recognition Method Based on Long-Term Voice Features and Fuzzy Logic

Speaker recognition depends on specific predefined steps. The most important steps are feature extraction and features matching. In addition, the category of the speaker voice features has an impact on the recognition process. The proposed speaker recognition makes use of biometric (voice) attributes to recognize the identity of the speaker. The long-term features were used such that maximum frequency, pitch and zero crossing rate (ZCR). In features matching step, the fuzzy inner product was used between feature vectors to compute the matching value between a claimed speaker voice utterance and test voice utterances. The experiments implemented using (ELSDSR) data set. These experiments showed that the recognition accuracy is 100% when using text dependent speaker recognition.

Text dependent Speaker Recognition using LPC_avg and Euclidean distance measure

Text dependent Speaker Recognition using LPC_avg and Euclidean distance measure

Speech Recognition Using Convolutional Neural Networks on Small Training Sets

The algorithm of separate words automatic recognition based on convolutional neural networks is developed and presented in this paper. Distinctive feature of this algorithm is the training on sets consisting of only hundreds or thousands of samples. Therefore, important problem is the selection of optimal architecture for neural network, which was firstly proposed and tested. After that, four different cases for recognition were researched: speaker-dependent recognition without noise, speaker-independent recognition without noise, speaker-dependent recognition with noise, speaker-independent recognition without noise. Finally, we analyse the experiment results that showed good results for all cases of interest.

Hearing impaired speech recognition: Stockwell features and models

The development of speech recognition system for recognising the speeches of a reasonable person in various languages usually is in fashion and does not involve challenges to be faced by the researchers. For the past ten years, advances are taking place in analysing and recognising the speeches of the hearing impaired because of the deployment of sophisticated processing methods to study the characteristics of speech production. These technological advancements not only pave the way for developing algorithms for recognising the speeches of healthy persons, also recognising the utterances of the hearing impaired. The study on analysing the oral communication skills of hearing-impaired children has received the attention of the researchers, speech pathologists and audiologists to develop assistive tool/system because inadequacy of such skills dramatically affects the social, educational and career opportunities available to them at large. This paper mainly emphasises the need for the development of a more challenging speaker independent speech recognition system for hearing impaired so that the system can respond to the speech uttered by any HI. In this work, Modified Group Delay Features and Stockwell transform cepstral features are used at the front end and vector quantisation (VQ) and Multivariate Hidden Markov Models (MHMM) at the back end for recognising the speeches uttered by any person with hearing disability. Performance of the system is compared for the three modelling techniques VQ, Fuzzy C Means (FCM) clustering and MHMM for the recognition of isolated digits in Tamil. Recognition accuracy is 89.25% and 79.5% for speaker dependent and independent speech recognition system for the hearing impaired. Performance of the system reveals that this system may be deployed to understand the speeches uttered by any hearing impaired speaker, improve the social status of the people with hearing impairment and mitigate the social stigma in leading a normal life.

Smart Autonomous Wheelchair Controlled by Voice Commands-Aided by Tracking System

This study focuses on the design of an autonomous wheelchair based smart driving features for disabled persons. The movement directions and position tracking of the wheelchair are controlled and localized by pre-defined voice commands and global positioning system (GPS), respectively. Arduino microcontroller based on speaker dependent voice recognition module and tracking system based on quad-board SIM808 has been used to help the wheelchair navigation. The experimental tests of the proposed system have been done and given satisfactory results in controlling the wheelchair and making a call on demand. Hence the proposed system is a simple, easy to use and lowcost hardware for designing.

A Cloud Assisted Hybrid Model Based Speaker Recognition and Resource Sharing

The technological world, with its current focus on ease and convenience, has undeniably shifted towards developing smart systems, smart buildings and cities. Machine learning and artificial intelligence have become vital tools used in the establishment of smart embedded systems. In this context, a cloud assisted hybrid model based speaker recognition system is proposed. It can be effectively integrated in smart spaces like buildings, offices, homes, etc. This paper proposes a multivariable hybrid model with speaker recognition as biometric verification for identity authentication. It is also connected to the cloud for further security and robustness. The system functionality is successfully tested. Preliminary results are presented. A higher subject dependent speaker recognition accuracy is achieved, for the studied case, with the devised solution.

Emotions recognition: different sets of features and models

The better and effective human–machine communication is ensured by performing affective computing. In the recent years, healthy research has been progressing in recognizing emotions by using various databases. This paper mainly emphasizes the effectiveness on the basis of using different sets of features and modeling techniques in evaluating the performance of multiple speaker-independent and speaker-dependent emotion recognition systems. It has become a challenging task to improve the performance of emotion recognition system, since EMO-DB Berlin database used in this work contains only ten speeches uttered by ten speakers in different emotions namely, Anger, Boredom, Disgust, Fear, Happiness, Sadness and Neutral. Speaker dependent and independent emotion recognition is done by creating models using clustering technique, Gaussian mixture modeling (GMM) and continuous density hidden Markov modeling (CDHMM) techniques for all emotions. The emotion recognition system is also evaluated for mel frequency cepstrum (MFCC) and concatenated MFCC with probability & shifted delta cepstrum (SDC), mel frequency linear predictive cepstrum (MFPLPC) and concatenated MFPLPC with probability & SDC and formants for clustering used as a modeling technique. These features provide complementary evidence in assessing the performance of the system based on VQ based clustering technique. This algorithm provides 99 and 100% as overall weighted accuracy recall (WAR) for performance evaluation with respect to correct identification of emotion for any one feature and modeling technique.

Speaker-Independent Silent Speech Recognition from Flesh-Point Articulatory Movements Using an LSTM Neural Network.

Silent speech recognition (SSR) converts non-audio information such as articulatory movements into text. SSR has the potential to enable persons with laryngectomy to communicate through natural spoken expression. Current SSR systems have largely relied on speaker-dependent recognition models. The high degree of variability in articulatory patterns across different speakers has been a barrier for developing effective speaker-independent SSR approaches. Speaker-independent SSR approaches, however, are critical for reducing the amount of training data required from each speaker. In this paper, we investigate speaker-independent SSR from the movements of flesh points on tongue and lip with articulatory normalization methods that reduce the inter-speaker variation. To minimize the across-speaker physiological differences of the articulators, we propose Procrustes matching-based articulatory normalization by removing locational, rotational, and scaling differences. To further normalize the articulatory data, we apply feature-space maximum likelihood linear regression and i-vector. In this paper, we adopt a bidirectional long short term memory recurrent neural network (BLSTM) as an articulatory model to effectively model the articulatory movements with long-range articulatory history. A silent speech data set with flesh points was collected using an electromagnetic articulograph (EMA) from twelve healthy and two laryngectomized English speakers. Experimental results showed the effectiveness of our speaker-independent SSR approaches on healthy as well as laryngectomy speakers. In addition, BLSTM outperformed standard deep neural network. The best performance was obtained by BLSTM with all the three normalization approaches combined.

Fractal dimension pattern-based multiresolution analysis for rough estimator of speaker-dependent audio emotion recognition

As a general means of expression, audio analysis and recognition have attracted much attention for its wide applications in real-life world. Audio emotion recognition (AER) attempts to understand the emotional states of human with the given utterance signals, and has been studied abroad for its further development on friendly human–machine interfaces. Though there have been several the-state-of-the-arts auditory methods devised to audio recognition, most of them focus on discriminative usage of acoustic features, while feedback efficiency of recognition demands is ignored. This makes possible application of AER, and rapid learning of emotion patterns is desired. In order to make predication of audio emotion possible, the speaker-dependent patterns of audio emotions are learned with multiresolution analysis, and fractal dimension (FD) features are calculated for acoustic feature extraction. Furthermore, it is able to efficiently learn the intrinsic characteristics of auditory emotions, while the utterance features are learned from FDs of each sub-band. Experimental results show the proposed method is able to provide comparative performance for AER.

Significance of incorporating excitation source parameters for improved emotion recognition from speech and electroglottographic signals

The work presented in this paper explores the effectiveness of incorporating the excitation source parameters such as strength of excitation and instantaneous fundamental frequency ( $$F_0$$ ) for emotion recognition task from speech and electroglottographic (EGG) signals. The strength of excitation (SoE) is an important parameter indicating the pressure with which glottis closes at the glottal closure instants (GCIs). The SoE is computed by the popular zero frequency filtering (ZFF) method which accurately estimates the glottal signal characteristics by attenuating or removing the high frequency vocaltract interactions in speech. The arbitrary impulse sequence, obtained from the estimated GCIs, is used to derive the instantaneous $$F_0$$ . The SoE and the instantaneous $$F_0$$ parameters are combined with the conventional mel frequency cepstral coefficients (MFCC) to improve the recognition rates of distinct emotions (Anger, Happy and Sad) using Gaussian mixture models as classifier. The performances of the proposed combination of SoE and instantaneous $$F_0$$ and their dynamic features with MFCC coefficients are compared with the emotion utterances (4 emotions and neutral) from classical German full blown emotion speech database (EmoDb) having simultaneous speech and EGG signals and Surrey Audio Visual Expressed Emotion database (3 emotions and neutral) for both speaker dependent and speaker independent emotion recognition scenarios. To reinforce the effectiveness of the proposed features and for better statistical consistency of the emotion analysis, a fairly large emotion speech database of 220 utterances per emotion in Tamil language with simultaneous EGG recordings, is used in addition to EmoDb. The effectiveness of SoE and instantaneous $$F_0$$ in characterizing different emotions is also confirmed by the improved emotion recognition performance in Tamil speech-EGG emotion database.

Arabic Isolated Word Speaker Dependent Recognition System

Arabic Isolated Word Speaker Dependent Recognition System

Design of Voice Controlled Smart Wheelchair

This paper describes the design of a smart, motorized, voice controlled wheelchair using embedded system. Proposed design supports voice activation system for physically disabled persons incorporating manual operation. Arduino microcontroller and speaker dependent voice recognition processor have been used to support the navigation of the wheel chair. The direction and velocity of the chair are controlled by pre-defined Arabic voice commands. The speaker dependent, isolated word recognition system (IWRS) for a definite utterance of Arabic words to suit the patient's requirements has been programmed and successfully demonstrated. The technique of speech signal processing for extraction of sound parameters, noise removal, intensity and normalization of time , and features matching etc. have been done with the speech processor HM2007that being embedded efficiently in real time. Arduino receives the coded digital signals from the IWRS which being properly recognizes voice commands in order to control the function of the chair accordingly. The wheelchair does not respond to a false speech command. The overall mechanical assembly is driven by two14A/24V/200Watt DC motors having an engagement/disengagement clutch and speed reduction gear with built-in locking control. The system is tested using a speech password to start operation and seven Arabic commands to control motion: Amam (forward), Saree'(fast), Batee' (slow), Khalf (backward), Yameen (right), Yesar (left), Tawaqaf (stop). It proved a good working in clear and noisy environments with safe movement.

Razvoj zbirke slovenskega čustvenega govora iz radijskih iger – EmoLUKS

V prispevku predstavljamo graditev slovenske zbirke čustvenega govora za umetno tvorjenje govora in hkrati raziščemo tudi možnosti njene uporabe pri razpoznavanju čustvenega stanja govorca. V prispevku se osredotočamo na opis razvite metodologije za označevanje paralingvistične informacije v govoru na primeru označevanja čustvenih stanj v slovenskih radijskih igrah. Zbirka vsebuje govorne zvočne signale sedemnajstih radijskih iger. Trenutno označeno gradivo obsega čustven govor enega govorca in ene govorke. Čustvene oznake posnetkov smo pridobili s pomočjo dvostopenjskega označevanja s petimi prostovoljnimi označevalci, ki so označili posnetke v dveh časovno ločenih intervalih. Način označevanja omogoča medsebojno primerjavo oznak označevalcev. S pomočjo označenega gradiva v obeh iteracijah poročamo o konsistentnosti označevalcev in ujemanju njihovih mnenj. Na podlagi večinskega mnenja pridobljenih čustvenih oznak vsakemu posnetku pripišemo tisto čustveno oznako, ki je bila med označevalci največkrat izbrana, in tako označene posnetke združimo v zbirko čustvenega govora EmoLUKS, ki jo kvantitativno in kvalitativno ovrednotimo z uporabo uveljavljenega samodejnega sistema za razpoznavanje čustvenih stanj govorca. Konsistentnost oznak ovrednotimo z dvorazrednim in sedemrazrednim od govorca odvisnim razvrščevalnikom čustvenih stanj. Uspešni rezultati razpoznavanja dodatno potrjujejo, da podatkovna zbirka kljub svoji zahtevnosti vsebuje jasno izražena čustvena stanja govorca.

English

This paper presents the effective and robust method for the feature extraction of the speaker dependent voice recognition. The authors developed a simple Matlab program for this purpose where the article discrete wavelet transform theory had been used. The voice of set of speakers had been inputted on the database and the discrete wavelet transform calculates the properties and variables needed in order to verify correctly the speaker. Experimental results show that our method is very effective and the results are satisfactory and finally, the wavelet-based voice recognition system and its performance are discussed and highlighted. Keywords—Speaker dependent, Voice recognition, Discrete wavelet transform every entry in an already stored code book database to recognize the voice said after the noise/silence removal, and (c) to use discrete wavelet transform (DWT) algorithm for the feature extraction for the recognition of a particular type of word spoken in the system. This paper is specifically limited to a speaker dependent voice recognition system and the method therein presented. The paper is organized as it follows: The section II briefly presents the fundamentals of the speech recognition, and discrete wavelet transform, Under the Section III experimental results are carried out in order to verify also the effectiveness of the proposed method. Conclusion ends the paper at Section IV.

Study of feature combination using HMM and SVM for multilingual Odiya speech emotion recognition

Emotions are broad aspects and expressed in a similar way by every human being; however, these are affected by culture. This creates a major threat to the universality of speech emotion detection system. Cultural behaviour of society affects the way emotions are expressed and perceived. Hence, an emotion recognition system customized for languages within a cultural group is feasible. In this work, a speaker dependent and speaker independent emotion recognition system has been proposed for two different dialects of Odisha: Sambalpuri and Cuttacki. Spectral speech features, such as, log power, Mel-frequency cepstral coefficients (MFCC), Delta MFCC, Double delta MFCC, log frequency power coefficients, and linear predictive cepstral coefficients, are used with Hidden Markov model and support vector machines (SVM) classifier, for classifying a speech into one of the seven discrete emotion classes: anger, happiness, disgust, fear, sadness, surprise, and neutral. For a better comparative study of system’s accuracy, features are taken individually as well as in combinations by varying sampling frequency, frame length and frame overlapping. Best average recognition accuracy obtained for speaker independent system, is 82.14 % for SVM classifier using only MFCC as feature vector. However, for speaker dependent system a hike in accuracy of more than 10 % is seen. It is also revealed that use of MFCC on SVM classifier, not only gives the best overall performance on 8 kHz sampling frequency, but also shows consistent performance for all the emotion classes, compared to other classifiers and feature combinations with less computational complexity. Hence, it can be applied efficiently in call centre application for emotion recognition over telephone.

Decoding Silent Speech in Japanese from Single Trial EEGS: Preliminary Results

We propose a new scheme for speaker-dependent silent speech recognition systems (SSRSs) using both single-trial electroencephalograms (EEGs) scalp-recorded and speech signals measured during overtly and covertly speaking “janken” and “season” in Japanese. This scheme consists of two phases. The learning phase specifies a Kalman filter using spectrograms of the speech signals and independent components (ICs), whose equivalent current dipole source localization (ECDL) solutions were located mainly at the Broca’s area, of the EEGs during the actual speech. In case of the “season” task, the speech signals were transformed into vowel and consonant sequences, and these relationships were learned by hidden Markov model (HMM) with Gaussian mixture densities. The decoding phase predicts spectrograms for the silent “janken” and “season” using the Kalman filter with the EEGs during the silent speech. For the silent “season”, the predicted spectrograms were inputted to the HMM, and which “season” was silently spoken was determined by the maximal log-likelihood among each HMM. Our preliminary results as training steps are as follows: the silent “jankens” were correctly discriminated; the silent “season”-HMMs worked well, suggesting that this scheme might be applied to the discrimination between all the pairs of the hiraganas