Voice Recognition Application Research Articles

A study on how to improve the accuracy of auto error correction from voice to text

Voice recognition application has been widely used in peoples daily lives. Usually, if people find it inconvenient to directly play the audio in public, or it is too noisy for them to listen to an audio message, they would probably use the voice recognition application in their mobile devices to translate the voice into text, so that they can understand the message clearly. However, there are still various errors and problems occurring during the usage of this common application, for instance, it can be hard sometimes to translate the sound into words correctly. This usually happens when the speaker speaks too fast and pronounces unclearly. Besides, some other factors, such as environmental noise, transmission channel quality, and the radio equipment, would also cause this problem. This paper mainly analyzes the causes of the inaccurate translation problem and some potential improvements to make this function more perfect. In conclusion, to solve this problem, the parity check matrix is a good way, since it can check whether the digits behind each word are still correct. After doing so, the digits will be changed into the correct words syntactically. However, even though the words are syntactically correct after applying the parity check matrix, they might not be semantically correct. Therefore, Levenshtin distance and Latent Semantic Analysis can be used to analyze the hidden meanings of words and sentences, so as to find the best suitable words to change.

Microcrack-assisted piezoelectric acoustic sensor based on f-MWCNTs/BaTiO3@PDMS nanocomposite and its self-powered voice recognition applications

Piezoelectric materials are suitable for use in wearable devices because of their simplicity and self-powering characteristics. However, before they can be used as wearable sensors, their flexibility and sensitivity need to be enhanced. Therefore, investing time and effort for improving piezoelectric materials is essential. In this study, we fabricated flexible piezoelectric nanogenerators (FPENGs) with microcracks based on a functionalized multi-walled carbon nanotubes/BaTiO3@polydimethylsiloxane composite. Then, we performed morphological and structural analyses of the fabricated composite. It was found that the presence of microcracks improved the output performance of the FPENG. The fabricated FPENG could function as an acoustic sensor with significantly enhanced frequency range sensitivity. As a proof-of-concept application, the ability of the FPENG to differentiate the loudness and melody of sound music, as well as its audio-to-text conversion capability, was demonstrated. These results validated the potential of the FPENG for a broader range of wireless communication applications on the Internet of Things.

Integration of an RSA-2048-bit public key cryptography solution in the development of secure voice recognition processing applications

The authors initially employs the fast Fourier transform (FFT) approach to transforming voice inputs into digital signals before integrating a speech recognition solution (which includes two models: the hidden Markov model (HMM) and the artificial neural network (ANN)). To achieve standard-tone identification of voice signals and digitally store speech, the authors then incorporated a 2048-bit Rivest-Shamir-Adleman (RSA) encryption method to encrypt and decrypt digital speech. The authors’ building team constructed the program using a 256-bit advanced encryption standard - Galois counter mode (AES-GCM) encryption method to assure the application’s effectiveness. The authors successfully created a voice recognition application according to the HMM of ANN. The collected findings suggest that the authors’ secure speech recognition program (named soft voice - RSA) has improved in terms of safety, keeping speech material secret, and speed. It takes roughly 0.2 s to generate a 2048-bit RSA key pair that exceeds the National Institute of Standards and Technology (NIST) standard, 700-1070 ms to process speech, 1-4 ms to encrypt 2048-bit RSA, 6-8 ms to decrypt 2048-bit RSA.

A dynamical convolutional neural network–based adaptive observer for information-poor systems

The objective of this paper is to propose a novel self-learning observer for effectively estimating unmeasured states of a dynamical information-poor system. By inspiring the success of convolutional neural network (CNN) in image and voice recognition applications, the dynamical convolutional neural network (DCNN) architecture (which is nonlinear-in-parameter unlike traditional basis functions–based neural network or fuzzy logic architectures) is proposed for approximating the unknown system dynamics, uncertainties, and disturbances. Instead of using an output observation error filtering method, complex weight-updating mechanism, strictly positive real (SPR) assumption, offline-tuning mechanisms, static approximation of gradients in back-propagation algorithm, or any other strong or strict requirement, the simple yet effective and novel rules for tuning DCNN weights and kernels are derived. Furthermore, the problems such as parameters over-tuning and unnecessary consumption of large amounts of computational power commonly found in traditional neuro-adaptive and fuzzy-adaptive controllers/observers are addressed and their solution is incorporated in the design of the observer. Thus, the proposed DCNN-based adaptive observer is energy-efficient. A quadratic function based on the convolution operator is also proposed which is utilized to provide the closed-loop stability analysis using Lyapunov’s approach. An example is simulated to demonstrate the performance of the proposed observer.

A VCO-Based ADC With Direct Connection to a Microphone MEMS, 80-dB Peak SNDR and 438-μW Power Consumption

This article presents a microphone readout chip incorporating a voltage controlled oscillator (VCO)-based ADC that can be directly connected to a capacitive micro electro mechanical systems (MEMSs) sensor without requiring a voltage buffer. The ADC uses an open-loop pseudo-differential architecture with two ring oscillators followed by a coarse–fine frequency-to-digital converter. The proposed coarse–fine architecture optimizes power consumption, thanks to a new algorithm. The MEMS is connected to the ring oscillators with a source follower circuit that can be programed to interchange signal to noise ratio (SNR) by the power consumption in small steps. This feature enables always-on operation in voice recognition applications. The microphone is compatible with standard PDM audio interfaces. Total ADC power consumption ranges between 245 and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$438 \mu \text{W}$ </tex-math></inline-formula> for peak signal-to-noise and distortion ratios (SNDRs) between 74 and 80.3 dB-A, including the ADC and MEMS coupling circuitry. The dynamic range (DR) achieves 108 dB at full performance with a total harmonic distortion (THD) of 1.5% at the acoustic overload point (AOP) of 128 dBSPL. The ADC occupies an area of 0.14 mm2 implemented in 0.13- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula> CMOS.

An On-Chip Fully Connected Neural Network Training Hardware Accelerator Based on Brain Float Point and Sparsity Awareness

In recent years, deep neural networks (DNNs) have brought revolutionary progress in various fields with the advent of technology. It is widely used in image pre-processing, image enhancement technology, face recognition, voice recognition, and other applications, gradually replacing traditional algorithms. It shows that the rise of neural networks has led to the reform of artificial intelligence. Since neural network algorithms are computationally intensive, they require GPUs or accelerated hardware for real-time computation. However, the high cost and high power consumption of GPUs result in low energy efficiency. It recently led to much research on accelerated digital circuit hardware design for deep neural networks. In this paper, we propose an efficient and flexible neural network training processor for fully connected layers. Our proposed training processor features low power consumption, high throughput, and high energy efficiency. It uses the sparsity of neuron activations to reduce the number of memory accesses and memory space to achieve an efficient training accelerator. The proposed processor uses a novel reconfigurable computing architecture to maintain high performance when operating Forward Propagation and Backward Propagation. The processor is implemented in Xilinx Zynq UltraSacle+MPSoC ZCU104 FPGA, with an operating frequency of 200MHz and power consumption of 6.444W, and can achieve 102.43 GOPS.

Human-Machine Convergence and Disruption of Socio-Cognitive Capabilities

People who have access to smartphone or computers are already using AI. A complex set of algorithm works in search engines, social networking sites, voice recognition applications, navigation applications, OTT platforms and video games. Human-machine convergence has been made possible because of the science and technology. Artificial Intelligence and Machine Learning are the contemporary technologies which is being used in several sectors for better precisions like medical diagnostics, remote sensing etc. There is a tremendous amount of potential for AI and ML to help individuals develop their mental and interpersonal skills. Since quite some time, researchers have been looking at what effect, if any, artificial intelligence may have on the decision-making process of humans. In the same way that it is used in the decision-making processes of the scientific and corporate communities, artificial intelligence is also being employed in the legal and gambling industries. Software and algorithms based on artificial intelligence (AI) have the potential to simulate human behaviour and predict its effects. Many people are of the opinion that artificial intelligence will improve the quality of human judgement and decision making, particularly in high-stakes circumstances like war. Investigating the structural, relational, and cognitive bases of artificial intelligence's function as a facilitator of decision-making is the goal of this line of study.&#x0D;

Ink-lithographic fabrication of silver-nanocrystal-based multiaxial strain gauge sensors through the coffee-ring effect for voice recognition applications

Human voice recognition techniques have remarkable potential for clinical applications because information from acoustic signals can reflect human body conditions. This paper reports the fabrication of Ag nanocrystal (NC)-based multiaxial wearable strain gauge sensors by ink-lithography for voice recognition systems. Benefiting from the one-step-device-fabrication strategy of ink-lithography, which can yield Ag NC patterns with specific dimensions and endow physical properties, the Ag NC-based multiaxial strain sensors can be fabricated on an ultrathin substrate (~ 6 μm). Additionally, the coffee-ring effect can be induced onto the Ag NC patterns to realize high sensitivity and angle dependence (gauge factors {G}_{0^circ } = 11.7 ± 1.2 and {G}_{90^circ } = 105.5 ± 20.1); moreover, the voice onset time for voice recognition can be detected by the sensors. These features assist in distinguishing between voiced and voiceless plosive contrasts via measurements of contact-based voice onset time differences and can act as a cornerstone for further advancements in wearable sensors as well as voice recognition and analysis.Graphical

IMPLEMENTASI METODE MEL-FREQUENCY CEPSTRAL COEFFICIENT DAN DTW PADA APLIKASI PENGENALAN SUARA TEMBANG SEKAR ALIT

The province of Bali has various types of songs that have different structures and functions, one of which is the Sekar Alit song. Seeing current technological advances, conservation efforts should follow the development of existing technology. This study recognizes the singer's voice with the MeliFrequency iCepstrum Coefficients (MFCC) method used to perform feature extraction, i.e. to obtain a parameter and information regarding the characteristics of a person's voice and the training of voice pattern matching against the voice of the DTW Singer Alit (Dinamic) and the Time Warping Alit method. . So that the final result of this research is an android-based tembang sekar alit learning application that can be used by students to learn more easily and effectively which is called the SekARAI application. The results of usability testing on the application found that the average value was above 3, which means that the SekARAI software that has been implemented has met the Usability element and besides that the software is easy to use and understand by users. The Sekar Alit Song Voice Recognition application has been evaluated using the confusion matrix method, it is found that the MFCC algorithm test results in an accuracy of 76.6%.

Classification of Arabic fricative consonants according to their places of articulation

&lt;span&gt;Many technology systems have used voice recognition applications to transcribe a speaker’s speech into text that can be used by these systems. One of the most complex tasks in speech identification is to know, which acoustic cues will be used to classify sounds. This study presents an approach for characterizing Arabic fricative consonants in two groups (sibilant and non-sibilant). From an acoustic point of view, our approach is based on the analysis of the energy distribution, in frequency bands, in a syllable of the consonant-vowel type. From a practical point of view, our technique has been implemented, in the MATLAB software, and tested on a corpus built in our laboratory. The results obtained show that the percentage energy distribution in a speech signal is a very powerful parameter in the classification of Arabic fricatives. We obtained an accuracy of 92% for non-sibilant consonants /f, χ, ɣ, ʕ, ћ, and h/, 84% for sibilants /s, sҁ, z, Ӡ and ∫/, and 89% for the whole classification rate. In comparison to other algorithms based on neural networks and support vector machines (SVM), our classification system was able to provide a higher classification rate.&lt;/span&gt;

Participation Action Research on Daily Health Literacy Using Voice Recognition Application for the Visual Impairment in Indonesia: A Research Protocol

Visually impaired people are often associated with various complex problems because they are considered unproductive even for daily health, so that daily health literacy is needed. The most needed aspects for visually impaired people are balanced nutrition, covid-19, and information on the nearest health service. The location of the difference in health literacy between people who see and people with visual impairment is accessibility. The purpose of this study is to describe the Participation Action Research (PAR) Protocol on Daily Health Literacy Using Voice Recognition Applications for visually impaired people in Indonesia. We will conduct a study using the Mixed Exploratory method, which is a method that in the early stages of research uses qualitative methods and the next stage uses quantitative methods. The mixing of the data of the two methods is connecting between the results of the first and subsequent studies. This exploratory type is a sequential model (sequence). The research stages are divided into three starting from pre-intervention by conducting systematic reviews, preliminary studies, and qualitative studies, as well as application design, the second stage is intervention through two groups, namely, the intervention group carried out health literacy using voice recognition applications and the control group carried out intervention using e-mail. -book reader, the intervention is continued with assistance for the blind by partners for 30 days, the third stage is post-intervention by monitoring and evaluating behavior, and evaluating applications for evolution and publication of applications to the play store for use by visually impaired people in Indonesia.

Hardware/Software Co-Design for TinyML Voice-Recognition Application on Resource Frugal Edge Devices

On-device artificial intelligence has attracted attention globally, and attempts to combine the internet of things and TinyML (machine learning) applications are increasing. Although most edge devices have limited resources, time and energy costs are important when running TinyML applications. In this paper, we propose a structure in which the part that preprocesses externally input data in the TinyML application is distributed to the hardware. These processes are performed using software in the microcontroller unit of an edge device. Furthermore, resistor–transistor logic, which perform not only windowing using the Hann function, but also acquire audio raw data, is added to the inter-integrated circuit sound module that collects audio data in the voice-recognition application. As a result of the experiment, the windowing function was excluded from the TinyML application of the embedded board. When the length of the hardware-implemented Hann window is 80 and the quantization degree is 2−5, the exclusion causes a decrease in the execution time of the front-end function and energy consumption by 8.06% and 3.27%, respectively.

The Application of Artificial Intelligence Voice Recognition on Helping Elders Use Mobile Phones More Easily

The problem that elders struggle to use mobile phone devices and elders often refuse to gain in touch with contacting devices with newest technologies have attracted people’s attention for a long time. It has already become a social phenomenon that elders tend to refuse smart devices such as smartphones. The reason is that most smartphones own a complex system, which has too high a difficulty for elders to master. The elders have a strong need for a way to remove the obstacle between their demands and the input of system orders. AI voice recognition, on the other hand, can basically remove this issue from consideration, for it can replace the manual manipulation part with machine-learning-based understanding of commands. Through deep learning methods mentioned in this paper, AI voice recognition can analyze orders given by voice, pick out key words and form into an order, and eventually execute the order properly. Voice recognition programs can also detect the elders’ health conditions through the attributes of the voice and make fast reactions automatically. This article aims to find a way to build targeted AI voice recognition systems for a specific kind of people, for example, the senior citizens, which is the target of the system in this article. This article also proposes rational prospects for future AI voice recognition applications in mental treatment or psychological researches areas.

Real time wearable speech recognition system for deaf persons

Numerous people around the world experience varying degrees of hearing difficulties. A sense of sound is critical for quality of life for people with hearing difficulties, including deaf people. Wearable devices have substantial potential for voice recognition applications owing to their low cost and light weight. They can assist deaf people perform their daily activities more easily without requiring assistance from others. I have designed and developed a wearable, vibration-based system that will enable deaf people to distinguish important sentences, thereby improving their quality of life. The device consists of a microphone and vibration motor and is mounted on the user’s back. A study was conducted with the device in which speech was detected using Mel Frequency Cepstral Coefficients (MFCC), pre-processing, and Dynamic Time Warping (DTW). Tests were conducted in three different signal-to-noise ratio environments in real time with a typical computer setup. The results had 95% accuracy, 97% specificity and 76% sensitivity.

Physical Activity Evaluation Using a Voice Recognition App: Development and Validation Study.

Historically, the evaluation of physical activity has involved a variety of methods such as the use of questionnaires, accelerometers, behavior records, and global positioning systems, each according to the purpose of the evaluation. The use of web-based physical activity evaluation systems has been proposed as an easy method for collecting physical activity data. Voice recognition technology not only eliminates the need for questionnaires during physical activity evaluation but also enables users to record their behavior without physically touching electronic devices. The use of a web-based voice recognition system might be an effective way to record physical activity and behavior. The purpose of this study was to develop a physical activity evaluation app to record behavior using voice recognition technology and to examine the app's validity by comparing data obtained using both the app and an accelerometer simultaneously. A total of 20 participants (14 men, 6 women; mean age 19.1 years, SD 0.9) wore a 3-axis accelerometer and inputted behavioral data into their smartphones for a period of 7 days. We developed a behavior-recording system with a voice recognition function using a voice recognition application programming interface. The exercise intensity was determined from the text data obtained by the voice recognition program. The measure of intensity was metabolic equivalents (METs). From the voice input data of the participants, 601 text-converted data could be confirmed, of which 471 (78.4%) could be automatically converted into behavioral words. In the time-matched analysis, the mean daily METs values measured by the app and the accelerometer were 1.64 (SD 0.20) and 1.63 (SD 0.20), respectively, between which there was no significant difference (P=.57). There was a significant correlation between the average METs obtained from the voice recognition app and the accelerometer in the time-matched analysis (r=0.830, P<.001). In the Bland-Altman plot for METs measured by the voice recognition app as compared with METs measured by accelerometer, the mean difference between the two methods was very small (0.02 METs), with 95% limits of agreement from -0.26 to 0.22 METs between the two methods. The average METs value measured by the voice recognition app was consistent with that measured by the 3-axis accelerometer and, thus, the data gathered by the two measurement methods showed a high correlation. The voice recognition method also demonstrated the ability of the system to measure the physical activity of a large number of people at the same time with less burden on the participants. Although there were still issues regarding the improvement of automatic text data classification technology and user input compliance, this research proposes a new method for evaluating physical activity using voice recognition technology.

Theoretical modelling and implementation of home energy management system using IoT based automation system

In this modern world a new technology is known as home automation has come into existence. Not only providing services for physically challenged people and elderly people but also it provides supporting system for the new modern world. The content in this paper summarizes the complete design which has been implemented. The Wireless Home Automation System which includes the application of voice recognition, makes this system simpler and easy to access. The next part of this paper includes foot step counter along with switching of light in automatic manner. The objective of this when the user has no time to install the application or if there is any error in connection of Bluetooth, the lights and other loads connected starts working when the person passes the door. Here the voice recognition application range has been observed as 1.5 m. We can achieve 85.25% accuracy in voice recognition along with four parameters include.

Voice Identification Using MFCC and Vector Quantization

The speaker identification is one of the fundamental problems in speech processing and voice modeling. The speaker identification applications include authentication in critical security systems and the accuracy of the selection. Large-scale voice recognition applications are a major challenge. Quick search in the speaker database requires fast, modern techniques and relies on artificial intelligence to achieve the desired results from the system. Many efforts are made to achieve this through the establishment of variable-based systems and the development of new methodologies for speaker identification. Speaker identification is the process of recognizing who is speaking using the characteristics extracted from the speech's waves like pitch, tone, and frequency. The speaker's models are created and saved in the system environment and used to verify the identity required by people accessing the systems, which allows access to various services that are controlled by voice, speaker identification involves two main parts: the first part is the feature extraction and the second part is the feature matching.

A Pilot Research on Android Based Voice Recognition Application

In recent trend, Speech recognition has become extensively used in customer service based organization. It has acquired great deal of research in pattern matching employed machine learning (learning speech by experience) and neural networks based speech endorsement domains. Speech recognition is the technology of capturing and perceiving human voice, interpreting it, producing text from it, managing digital devices and assisting visually impaired and older adults using unequivocal digital signal processing. In this paper we have presented a comprehensive study of different methodologies in android enabled speech recognition system that focused at analysis of the operability and reliability of voice note app. Subsequently we have suggested and experimented an android based speech recognizer app viz. Annotate which predominately focus on voice dictation in five different languages (English, Hindi, Tamil, Malayalam and Telugu) and extracting text from image using Automatic Speech Recognition (ASR) and Optical Character Recognition (OCR) algorithm. Finally, we identified opportunities for future enhancements in this realm.

DGR: Gender Recognition of Human Speech Using One-Dimensional Conventional Neural Network

The speech entailed in human voice comprises essentially paralinguistic information used in many voice-recognition applications. Gender voice is considered one of the pivotal parts to be detected from a given voice, a task that involves certain complications. In order to distinguish gender from a voice signal, a set of techniques have been employed to determine relevant features to be utilized for building a model from a training set. This model is useful for determining the gender (i.e., male or female) from a voice signal. The contributions are three-fold including (i) providing analysis information about well-known voice signal features using a prominent dataset, (ii) studying various machine learning models of different theoretical families to classify the voice gender, and (iii) using three prominent feature selection algorithms to find promisingly optimal features for improving classification models. The experimental results show the importance of subfeatures over others, which are vital for enhancing the efficiency of classification models’ performance. Experimentation reveals that the best recall value is equal to 99.97%; the best recall value is 99.7% for two models of deep learning (DL) and support vector machine (SVM), and with feature selection, the best recall value is 100% for SVM techniques.

Stretchable Micromotion Sensor with Enhanced Sensitivity Using Serpentine Layout.

The application of the serpentine mesh layout in stretchable electronics provides a feasible method to achieve the desired stretchability by structural design instead of modifying the intrinsic mechanical properties of the applied materials. However, previous worksusing the serpentine layout mainly focused on the optimization of structural stretchability. In this paper, the serpentine mesh design concept is used to transform the high-performance but hard-to-stretch piezoelectric film into a stretchable form. The serpentine layout design strategies for the piezoelectric film, which aim at not only desired stretchability but also high utilization of the strain in the piezoelectric film during deformation, are discussed with experimental and computational results. A stretchable micromotion sensor with high sensitivity is realized using the piezoelectric film with a serpentine layout. Human voice recognition applications of the sensor, including speech pattern recognition with machine learning, are demonstrated with the sensor integrated with a wireless module. The stretchable micromotion sensor with a serpentine layout illustrates the broader application of serpentine layout design in the functional materials of stretchable electronics, which can further extend the range of available functional materials for novel stretchable electronic devices.