Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “SPEECH”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 271 records · Page 15

Voice stress analysis

In a study of the validity of eight candidate voice measures (fundamental frequency, amplitude, speech rate, frequency jitter, amplitude shimmer, Psychological Stress Evaluator scores, energy distribution, and the derived measure of the above measures) for determining psychological stress, 17 males age 21 to 35 were subjected to a tracking task on a microcomputer CRT while parameters of vocal production as well as heart rate were measured. Findings confirm those of earlier studies that increases in fundamental frequency, amplitude, and speech rate are found in speakers involved in extreme levels of stress. In addition, it was found that the same changes appear to occur in a regular fashion within a more subtle level of stress that may be characteristic, for example, of routine flying situations. None of the individual speech measures performed as robustly as did heart rate.

Brenner, Malcolm↗

A 4.8 kbps code-excited linear predictive coder

A secure voice system STU-3 capable of providing end-to-end secure voice communications (1984) was developed. The terminal for the new system will be built around the standard LPC-10 voice processor algorithm. The performance of the present STU-3 processor is considered to be good, its response to nonspeech sounds such as whistles, coughs and impulse-like noises may not be completely acceptable. Speech in noisy environments also causes problems with the LPC-10 voice algorithm. In addition, there is always a demand for something better. It is hoped that LPC-10's 2.4 kbps voice performance will be complemented with a very high quality speech coder operating at a higher data rate. This new coder is one of a number of candidate algorithms being considered for an upgraded version of the STU-3 in late 1989. The problems of designing a code-excited linear predictive (CELP) coder to provide very high quality speech at a 4.8 kbps data rate that can be implemented on today's hardware are considered.

Tremain, Thomas E.↗

Sinusoidal transform coding

It has been shown that an analysis/synthesis system based on a sinusoidal representation of speech leads to synthetic speech that is essentially perceptually indistinguishable from the original. Strategies for coding the amplitudes, frequencies and phases of the sine waves have been developed that have led to a multirate coder operating at rates from 2400 to 9600 bps. The encoded speech is highly intelligible at all rates with a uniformly improving quality as the data rate is increased. A real-time fixed-point implementation has been developed using two ADSP2100 DSP chips. The methods used for coding and quantizing the sine-wave parameters for operation at the various frame rates are described.

Mcaulay, Robert J.↗

Laboratory and in-flight experiments to evaluate 3-D audio display technology

Laboratory and in-flight experiments were conducted to evaluate 3-D audio display technology for cockpit applications. A 3-D audio display generator was developed which digitally encodes naturally occurring direction information onto any audio signal and presents the binaural sound over headphones. The acoustic image is stabilized for head movement by use of an electromagnetic head-tracking device. In the laboratory, a 3-D audio display generator was used to spatially separate competing speech messages to improve the intelligibility of each message. Up to a 25 percent improvement in intelligibility was measured for spatially separated speech at high ambient noise levels (115 dB SPL). During the in-flight experiments, pilots reported that spatial separation of speech communications provided a noticeable improvement in intelligibility. The use of 3-D audio for target acquisition was also investigated. In the laboratory, 3-D audio enabled the acquisition of visual targets in about two seconds average response time at 17 degrees accuracy. During the in-flight experiments, pilots correctly identified ground targets 50, 75, and 100 percent of the time at separation angles of 12, 20, and 35 degrees, respectively. In general, pilot performance in the field with the 3-D audio display generator was as expected, based on data from laboratory experiments.

Ericson, Mark↗

Signal Prediction With Input Identification

A novel coding technique is presented for signal prediction with applications including speech coding, system identification, and estimation of input excitation. The approach is based on the blind equalization method for speech signal processing in conjunction with the geometric subspace projection theory to formulate the basic prediction equation. The speech-coding problem is often divided into two parts, a linear prediction model and excitation input. The parameter coefficients of the linear predictor and the input excitation are solved simultaneously and recursively by a conventional recursive least-squares algorithm. The excitation input is computed by coding all possible outcomes into a binary codebook. The coefficients of the linear predictor and excitation, and the index of the codebook can then be used to represent the signal. In addition, a variable-frame concept is proposed to block the same excitation signal in sequence in order to reduce the storage size and increase the transmission rate. The results of this work can be easily extended to the problem of disturbance identification. The basic principles are outlined in this report and differences from other existing methods are discussed. Simulations are included to demonstrate the proposed method.

Juang, Jer-Nan↗

Talking Wheelchair

Communication is made possible for disabled individuals by means of an electronic system, developed at Stanford University's School of Medicine, which produces highly intelligible synthesized speech. Familiarly known as the "talking wheelchair" and formally as the Versatile Portable Speech Prosthesis (VPSP). Wheelchair mounted system consists of a word processor, a video screen, a voice synthesizer and a computer program which instructs the synthesizer how to produce intelligible sounds in response to user commands. Computer's memory contains 925 words plus a number of common phrases and questions. Memory can also store several thousand other words of the user's choice. Message units are selected by operating a simple switch, joystick or keyboard. Completed message appears on the video screen, then user activates speech synthesizer, which generates a voice with a somewhat mechanical tone. With the keyboard, an experienced user can construct messages as rapidly as 30 words per minute.

Source record↗

Field-testing the new DECtalk PC system for medical applications

Synthesized human speech has now reached a new level of performance. With the introduction of DEC's new DECtalk PC, the small system developer will have a very powerful tool for creative design. It has been our privilege to be involved in the beta-testing of this new device and to add a medical dictionary which covers a wide range of medical terminology. With the inherent board level understanding of speech synthesis and the medical dictionary, it is now possible to provide full digital speech output for all medical files and terms. The application of these tools will cover a wide range of options for the future and allow a new dimension in dealing with the complex user interface experienced in medical practice.

NASA Discipline Data Analysis↗

Communication as group process media of aircrew performance

This study of group process was motivated by a high-fidelity flight simulator project in which aircrew performance was found to be better when the crew had recently flown together. Considering recent operating experience as a group-level input factor, aspects of the communication process between crewmembers (Captain and First Officer), were explored as a possible mediator to performance. Communication patterns were defined by a speech act typology adapted for the flightdeck setting and distinguished crews that had previously flown together (FT) from those that had not flown together (NFT). A more open communication channel with respect to information exchange and validation and greater First Officer participation in task-related topics was shown by FT crews while NFT crews engaged in more non-task discourse, a speech mode less structured by roles and probably serving a more interpersonal function. Relationships between the speech categories themselves, representing linguistic, and role-related interdependencies provide guidelines for interpreting the primary findings.

Aviation↗

Spoken Language Processing in the Clarissa Procedure Browser

Clarissa, an experimental voice enabled procedure browser that has recently been deployed on the International Space Station, is as far as we know the first spoken dialog system in space. We describe the objectives of the Clarissa project and the system's architecture. In particular, we focus on three key problems: grammar-based speech recognition using the Regulus toolkit; methods for open mic speech recognition; and robust side-effect free dialogue management for handling undos, corrections and confirmations. We first describe the grammar-based recogniser we have build using Regulus, and report experiments where we compare it against a class N-gram recogniser trained off the same 3297 utterance dataset. We obtained a 15% relative improvement in WER and a 37% improvement in semantic error rate. The grammar-based recogniser moreover outperforms the class N-gram version for utterances of all lengths from 1 to 9 words inclusive. The central problem in building an open-mic speech recognition system is being able to distinguish between commands directed at the system, and other material (cross-talk), which should be rejected. Most spoken dialogue systems make the accept/reject decision by applying a threshold to the recognition confidence score. NASA shows how a simple and general method, based on standard approaches to document classification using Support Vector Machines, can give substantially better performance, and report experiments showing a relative reduction in the task-level error rate by about 25% compared to the baseline confidence threshold method. Finally, we describe a general side-effect free dialogue management architecture that we have implemented in Clarissa, which extends the "update semantics'' framework by including task as well as dialogue information in the information state. We show that this enables elegant treatments of several dialogue management problems, including corrections, confirmations, querying of the environment, and regression testing.

Rayner, M.↗

Communication system with adaptive noise suppression

A signal-to-noise ratio dependent adaptive spectral subtraction process eliminates noise from noise-corrupted speech signals. The process first pre-emphasizes the frequency components of the input sound signal which contain the consonant information in human speech. Next, a signal-to-noise ratio is determined and a spectral subtraction proportion adjusted appropriately. After spectral subtraction, low amplitude signals can be squelched. A single microphone is used to obtain both the noise-corrupted speech and the average noise estimate. This is done by determining if the frame of data being sampled is a voiced or unvoiced frame. During unvoiced frames an estimate of the noise is obtained. A running average of the noise is used to approximate the expected value of the noise. Spectral subtraction may be performed on a composite noise-corrupted signal, or upon individual sub-bands of the noise-corrupted signal. Pre-averaging of the input signal's magnitude spectrum over multiple time frames may be performed to reduce musical noise.

Kozel, David↗

Noise-Canceling Helmet Audio System

A prototype helmet audio system has been developed to improve voice communication for the wearer in a noisy environment. The system was originally intended to be used in a space suit, wherein noise generated by airflow of the spacesuit life-support system can make it difficult for remote listeners to understand the astronaut s speech and can interfere with the astronaut s attempt to issue vocal commands to a voice-controlled robot. The system could be adapted to terrestrial use in helmets of protective suits that are typically worn in noisy settings: examples include biohazard, fire, rescue, and diving suits. The system (see figure) includes an array of microphones and small loudspeakers mounted at fixed positions in a helmet, amplifiers and signal-routing circuitry, and a commercial digital signal processor (DSP). Notwithstanding the fixed positions of the microphones and loudspeakers, the system can accommodate itself to any normal motion of the wearer s head within the helmet. The system operates in conjunction with a radio transceiver. An audio signal arriving via the transceiver intended to be heard by the wearer is adjusted in volume and otherwise conditioned and sent to the loudspeakers. The wearer s speech is collected by the microphones, the outputs of which are logically combined (phased) so as to form a microphone- array directional sensitivity pattern that discriminates in favor of sounds coming from vicinity of the wearer s mouth and against sounds coming from elsewhere. In the DSP, digitized samples of the microphone outputs are processed to filter out airflow noise and to eliminate feedback from the loudspeakers to the microphones. The resulting conditioned version of the wearer s speech signal is sent to the transceiver.

Seibert, Marc A.↗

Natural Language Based Multimodal Interface for UAV Mission Planning

As the number of viable applications for unmanned aerial vehicle (UAV) systems increases at an exponential rate, interfaces that reduce the reliance on highly skilled engineers and pilots must be developed. Recent work aims to make use of common human communication modalities such as speech and gesture. This paper explores a multimodal natural language interface that uses a combination of speech and gesture input modalities to build complex UAV flight paths by defining trajectory segment primitives. Gesture inputs are used to define the general shape of a segment while speech inputs provide additional geometric information needed to fully characterize a trajectory segment. A user study is conducted in order to evaluate the efficacy of the multimodal interface.

Chandarana, Meghan↗

Machine Learning for Predicting Team Functioning in HERA Missions

Team functioning is integral to success in future long term space exploration missions. Proactively detecting declines in team functioning can mitigate conflict and ensure mission success. This project developed a speech-based artificial intelligence (AI) system that unobtrusively predicts degradation in team functioning, including performance and cohesion, in the Human Exploration Research Analog (HERA) Campaigns 4 and 5. The AI system conducted automated analysis of the prosodic (tone of voice) and linguistic (language content) components of speech, modeling interpersonal dynamics at both the turn-taking and day-wide levels. We investigated team functioning via observing structured interactions (i.e., multi-mission space exploration vehicle-extra vehicular activity [MMSEV-EVA], team interaction battery [TIB]) and unstructured interactions before the MMSEV-EVA task. We developed machine learning models to predict team functioning (objective task accuracy, self reported team efficacy and self reported team cohesion) by analyzing OpenSmile acoustic features, linguistic descriptors extracted via the linguistic inquiry and word count (LIWC) dictionary, and semantic embeddings. In the TIB, static models using logistic regression and random forests were not able to predict task accuracy, but predicted team efficacy and cohesion during both the decision making and relational tasks to a moderate level (60-70%). Majority voting on the individual turns to predict day long team efficacy further increased accuracies (70-80%). Finally, long short-term memory (LSTM) models showed the best performance across all variables (80-91%), including task performance. In the MMSEV-EVA, static models achieved an accuracy of 60% with majority voting, which increased to 80% through the incorporation of mission day as a variable, accounting for the learning effect. A key finding across both tasks was the "team-dependent" nature of these interactions; models achieved much higher accuracy when trained on prior days of the same team's data rather than attempting to generalize across entirely different teams, with even 1-2 days of prior data per team achieving 5-15% improvement over team-independent models. In addition, the incorporation of pre-task data from the same team also improves model performance, e.g., incorporating data from the decision-making task of the TIB, which preceded the relational task, improved the prediction of team efficacy and cohesion during the latter. We compared model performance when trained on machine-generated data compared to data that had been further corrected by human annotators. Overall, models trained on human-corrected data exhibited a modest improvement in performance, particularly when acoustic features were used. We found no significant correlation between word error rate (WER) and model accuracy (r(55) = -0.08, p = 0.51), but model’s accuracy was significantly higher for medium/high quality transcription (0.74 (SD = 0.48)) compared to the low-quality group (0.64 (SD = 0.36)) (t(63)=2.82, p = 0.006). Based on these, several design recommendation emerge, that could inform Standards at NASA. Models predicting team functioning should incorporate at least one to two days of historical interaction data, include brief pre-task discussions, and explicitly model temporal learning effects, especially for longer operational tasks. Minimum quality standards for automated speech-processing pipelines are needed, given the performance gains observed with manually corrected acoustic data. Finally, systems should leverage both acoustic features and language embeddings in complementary ways, with modality choices and fusion strategies tailored to mission context, task demands, and data quality requirements.

Shrivatsa Mishra↗

RADIOISOTOPE IDENTIFICATION OF SHIELDED AND MASKED SNM RDD MATERIALS

Sonar and speech techniques have been investigated to improve functionality and enable handheld and other man-portable, mobile, and portal systems to positively detect and identify illicit nuclear materials, with minimal data and with minimal false positives and false negatives. RadSonar isotope detection and identification is an algorithm development project funded by NA-22 and employing the resources of Savannah River National Laboratory and three University Laboratories (JHU-APL, UT-ARL, and UW-APL). Algorithms have been developed that improve the probability of detection and decrease the number of false positives and negatives. Two algorithms have been developed and tested. The first algorithm uses support vector machine (SVM) classifiers to determine the most prevalent nuclide(s) in a spectrum. It then uses a constrained weighted least squares fit to estimate and remove the contribution of these nuclide(s) to the spectrum, iterating classification and fitting until there is nothing of significance left. If any Special Nuclear Materials (SNMs) were detected in this process, a second tier of more stringent classifiers are used to make the final SNM alert decision. The second algorithm is looking at identifying existing feature sets that would be relevant in the radioisotope identification context. The underlying philosophy here is to identify parallels between the physics and/or the structures present in the data for the two applications (speech analysis and gamma spectroscopy). The expectation is that similar approaches may work in both cases. The mel-frequency cepstral representation of spectra is widely used in speech, particularly for two reasons: approximation of the response of the human ear, and simplicity of channel effect separation (in this context, a 'channel' is a method of signal transport that affects the signal, examples being vocal tract shape, room echoes, and microphone response). Measured and simulated gamma-ray spectra from a hand-held Radioisotope Identification Device were used to evaluate the algorithms. This paper will present and discuss results of the Test and Evaluation performed on two algorithms produced from the project.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Many but not all deep neural network audio models capture brain responses and exhibit correspondence between model stages and brain regions

Models that predict brain responses to stimuli provide one measure of understanding of a sensory system and have many potential applications in science and engineering. Deep artificial neural networks have emerged as the leading such predictive models of the visual system but are less explored in audition. Prior work provided examples of audio-trained neural networks that produced good predictions of auditory cortical fMRI responses and exhibited correspondence between model stages and brain regions, but left it unclear whether these results generalize to other neural network models and, thus, how to further improve models in this domain. We evaluated model-brain correspondence for publicly available audio neural network models along with in-house models trained on 4 different tasks. Most tested models outpredicted standard spectromporal filter-bank models of auditory cortex and exhibited systematic model-brain correspondence: Middle stages best predicted primary auditory cortex, while deep stages best predicted non-primary cortex. However, some state-of-the-art models produced substantially worse brain predictions. Models trained to recognize speech in background noise produced better brain predictions than models trained to recognize speech in quiet, potentially because hearing in noise imposes constraints on biological auditory representations. The training task influenced the prediction quality for specific cortical tuning properties, with best overall predictions resulting from models trained on multiple tasks. The results generally support the promise of deep neural networks as models of audition, though they also indicate that current models do not explain auditory cortical responses in their entirety.

59 BASIC BIOLOGICAL SCIENCES↗

Human Sensorimotor Cortex Control of Directly Measured Vocal Tract Movements during Vowel Production

During speech production, we make vocal tract movements with remarkable precision and speed. Our understanding of how the human brain achieves such proficient control is limited, in part due to the challenge of simultaneously acquiring high-resolution neural recordings and detailed vocal tract measurements. To overcome this challenge, we combined ultrasound and video monitoring of the supralaryngeal articulators (lips, jaw, and tongue) with electrocorticographic recordings from the cortical surface of 4 subjects (3 female, 1 male) to investigate how neural activity in the ventral sensory-motor cortex (vSMC) relates to measured articulator movement kinematics (position, speed, velocity, acceleration) during the production of English vowels. We found that high-gamma activity at many individual vSMC electrodes strongly encoded the kinematics of one or more articulators, but less so for vowel formants and vowel identity. Neural population decoding methods further revealed the structure of kinematic features that distinguish vowels. Encoding of articulator kinematics was sparsely distributed across time and primarily occurred during the time of vowel onset and offset. In contrast, encoding was low during the steady-state portion of the vowel, despite sustained neural activity at some electrodes. Significant representations were found for all kinematic parameters, but speed was the most robust. These findings enabled by direct vocal tract monitoring demonstrate novel insights into the representation of articulatory kinematic parameters encoded in the vSMC during speech production.

47 OTHER INSTRUMENTATION↗

Effects of Voice Rehabilitation After Radiation Therapy for Laryngeal Cancer: A Randomized Controlled Study

Background: Patients treated with radiation therapy for laryngeal cancer often experience voice problems. The aim of this randomized controlled trial was to assess the efficacy of voice rehabilitation for laryngeal cancer patients after having undergone radiation therapy and to investigate whether differences between different tumor localizations with regard to rehabilitation outcomes exist. Methods and Materials: Sixty-nine male patients irradiated for laryngeal cancer participated. Voice recordings and self-assessments of communicative dysfunction were performed 1 and 6 months after radiation therapy. Thirty-three patients were randomized to structured voice rehabilitation with a speech-language pathologist and 36 to a control group. Furthermore, comparisons with 23 healthy control individuals were made. Acoustic analyses were performed for all patients, including the healthy control individuals. The Swedish version of the Self Evaluation of Communication Experiences after Laryngeal Cancer and self-ratings of voice function were used to assess vocal and communicative function. Results: The patients who received vocal rehabilitation experienced improved self-rated vocal function after rehabilitation. Patients with supraglottic tumors who received voice rehabilitation had statistically significant improvements in voice quality and self-rated vocal function, whereas the control group did not. Conclusion: Voice rehabilitation for male patients with laryngeal cancer is efficacious regarding patient-reported outcome measurements. The patients experienced better voice function after rehabilitation. Patients with supraglottic tumors also showed an improvement in terms of acoustic voice outcomes. Rehabilitation with a speech-language pathologist is recommended for laryngeal cancer patients after radiation therapy, particularly for patients with supraglottic tumors.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Human Systems Integration Approach in Implementing Voice-Control of Future Spacecraft Systems

Crewed spacecraft and habitats of the future will require more automation and autonomy to support complex missions. However, these complex systems require a more efficient command and control input method. Speech recognition along with visual or auditory feedback is an alternative, providing an extra pair of hands and eyes for the crew. Yet, speech recognition demands a highly integrated development approach to ensure a successful system implementation. To ensure the voice control application is developed correctly will require a Human Systems Integration (HSI) approach. This paper provides an insight into the development of a speech/voice control application of a spacecraft system that encompasses automation and autonomy through an HSI approach. Results of the voice control experiment of the Space Shuttle camera system are provided as lessons learned about voice control on a spacecraft. Limitations and challenges of the technology are addressed as well as how HSI can help develop these types of voice control command and control systems.

Voice control↗