Search

When intelligibility is measured, objective estimation is more convenient than subjective. However, most objective estimation methods estimate monaural intelligibility using monaural signals. Therefore, it is necessary to estimate binaural intelligibility using binaural signals in order to take into account that a person listens using two ears. We evaluated an intelligibility estimation method using the Better Ear Model which selects better value out of the left and right feature values, and found that high estimation accuracy is possible. Accordingly, we proposed and evaluated an extension to this model which divides the signal into critical frequency bands, and takes the better value between left and right channel in each band (Band-Selection Better Ear Model). Furthermore, we tried estimation using machine learning in addition to regression analysis which we applied previously. Neural network and support vector machine were used here. Comparison between the Better Ear Model and the Band-Selection Bette...

Subjective speech intelligibility assessment is important for the development of outdoor public address system. However, as this assessment is difficult in many cases, we propose an objective speech intelligibility evaluation system that includes a machine learning technique. In this talk, we have proved a subjective evaluation and objective prediction of speech intelligibility using the outdoor public address systems at 10 locations in Sendai City, where impulse responses were recorded after the Great East Japan Earthquake. First, the results of the subjective intelligibility evaluation by different test word lists with the same sound field conditions showed that the root mean squared error (RMSE) was very small, not exceeding 7.0%. Next, we generated the intelligibility prediction model trained with true/false results of 22 subjects using the support vector machine (SVM). This prediction model extracted the feature vector, using the ITU-T P.563 speech quality feature set of the test speech signal. We ev...

This paper describes initial results of speech source direction identification experiments on the horizontal plane. This is an initial study for the sound localization required in order to separate the individual speech source location to increase the intelligibility in remote conference systems. Our focus here was to determine the effect of sound source characteristics on the source direction identification accuracy. Sound sources were located on three target locations (forward, right, and left) and five (45 deg in both right and left added) on the horizontal plane using a real‐time virtual acoustic environment rendering system (NASA SLAB). SLAB integrates HRTFs to render the source localization. The sound sources used were conversation speech, white noise, and conversation speech mixed with noise at an SNR of 0 dB. The observers were asked to identify the direction of the rendered sound presented over a headphone. The results showed that the direction identification accuracy of conversational speech is ...

This paper describes a three‐dimensional (3‐D) conference system using an open source software library on conventional PCs. We will attempt to use both 3‐D graphics and audio to construct a virtual conference environment for effective communication between remote parties. A rough prototype system was developed using OpenGL and OpenAL. The system uses local files for voice output, whose image location is rendered according to the user input. We initially evaluated the perceived sound image location accuracy of the rendered sound image using the prototype. Users were asked to identify the location of the rendered sound image from among four choices: front, back, left, and right. The users were able to identify the left and right images correctly at virtually 100%, but the front and back identification were lower than 10% for some sounds, particularly male speech. We plan to implement audio‐streaming functions to achieve real‐time audio conferencing and evaluate the benefits of 3‐D audio for conferences. We ...

We investigated the use of objective quality measures to estimate the intelligibility of Japanese speech. We initially focused on PESQ (perceptual evaluation of speech quality), which is the state‐of‐the‐art objective assessment method, and can estimate the mean opinion scores (MOS) at an extremely high accuracy. Since we can assume that speech quality is correlated with intelligibility, it should be possible to estimate the intelligibility from the estimated opinion scores or some of its derivatives. We tried to estimate the intelligibility of the Japanese Rhyme Test (DRT). The DRT uses minimal word pairs whose initial phone differs by one phonetic feature. We estimated the MOS of the DRT word samples mixed with noise and tried to correlate this with the measured intelligibility. However, the estimated MOS showed no difference between phonetic features. However, the difference in the estimated MOS between the word pairs seems to differ by a phonetic feature for SNR above 10 dB, which suggests that the estimation of intelligibility by a phonetic feature may be possible. We also plan to selectively use the internal data used to calculate the MOS estimate for better intelligibility estimation.

A speech packet communication system for making speech communication among a plurality of speech terminals and including a communication network. The communication network includes a transmission path and a plurality of intermediate switching nodes, each controlling the traffic of a plurality of speech packets received from the transmission path or other intermediate switching nodes in accordance with priority indicators contained in the respective packets. Each speech terminal converts the input speech signal to packets, transmits the packets onto the transmission path or decodes speech packets received from the transmission path. Particularly, when the terminal converts the input speech signal to packets, it uses a special purpose function of determining a priority indicator assigned to the packet to greatly improve the quality of a reproduced speech at a receiving end.

Phonetic modeling includes the steps of forming triphone grammars ( 11 ) from phonetic data, training triphone models ( 13 ), clustering triphones ( 14 ) that are acoustically close together and mapping unclustered triphone grammars into a clustered model ( 16 ). The clustering process includes using a decision tree based on the acoustic likelihood and allows sub-model clusters in user-definable units.

Improved speech recognition is achieved according to the present invention by use of between word and/or between phrase coarticulation. The increase in the number of phonetic models required to model this additional vocabulary is reduced by clustering 19, 20 the inter-word/phrase models and grammar into only a few classes. By using one class for consonant inter-word context and two classes for vowel contexts, the accuracy for Japanese was almost as good as for unclustered models while the number of models was reduced more than half.

In a speech packet transmission system, a transmitter (speech terminal) first linearly predicts a speech signal input at a predetermined sample period from the input signal in the past and outputs the difference between the input speech signal and the prediction signal. A prediction coefficient used for linear prediction at this time is adapted with an output differential coded value at each predetermined sample period such that the prediction value is as close as possible to the value of the input waveform. The difference between the input speech signal and the prediction signal is converted to a differential code. A packet is prepared from a plurality of differential codes obtained during a frame period concerned and transmitted to a transmission line. A bit accuracy indicative of the differential code and a bit accuracy indicative of the differential coded value used for adaptation of the prediction coefficient are changed depending on the nature or property of the input speech signal.

A speech signal is analyzed for each frame so that it is separated into spectral envelope information and excitation information, and the excitation information is expressed by a plurality of pulses. Judgement is conducted as to whether the current frame is a voiced frame immediately after the transition from an unvoiced frame, a voiced frame continuative from a voiced frame or an unvoiced frame, and excitation pulses are generated in accordance with the judgement result. In case of a continuing voiced frame, the excitation pulse position of the current voiced frame is determined based on the pitch period with respect to the excitation pulse position of the immediately preceding voiced frame so that the excitation pulse train is generated at a position approximated to the determined position.

A system for and a method of multiplexing digital speech signals corresponding to speech signals inputted from a plurality of terminals, the digital speech signals including a plurality of speech unit signals, in which a weighted priority signal is produced for the speech unit signal of each of the speech signals from the terminals, an assignement signal is produced from the priority signals indicative of which of the speech unit signals be selected for transmission in the form of a train of multiplexed speech unit signals in a cyclic transmission period of time, so that the speech unit signals are multiplexed on the basis of the assignment signal.

In a speech coding method and system in which a speech signal is analyzed in each frame so as to be separated into spectral envelope information and excitation information and both of the information are coded, each frame is divided into a plurality of sub-frames and a pulse of the maximum-amplitude is extracted from pulses within each sub-frame in order to provide large-amplitude pulses from each frame, thereby greatly reducing the number of pulse extracting processing steps.

A speech coding system includes apparatus for generating a variable threshold dependent upon the power of an input speech signal, and a comparator for comparing the power of the input speech signal with the variable threshold value to generate a discriminating signal for discriminating between a period when a speech continues and a period when the speech pauses, to change the coding operation for the input speech signal in accordance with the level of the discriminating signal, thereby forming voiced and unvoiced frames independently of each other.

A speech signal transmission system for transmitting coded speech signals with variable bit rate is disclosed, which comprises a coder for analyzing digital speech signals inputted in a one-frame period and transforming them into coded data comprising a plurality of parameters indicating characteristics of the inputted speech signals, a data arranging circuit for arranging the coded data outputted by the coder in the order of the priority in the decoding of the speech signals and outputting them, and a bit stealer allowing a series of coded data outputted successively by the data arranging circuit to pass through only in a period of time determined by the transmission bit rate. The bit arranging circuit outputs the parameters in the order of the priority of decomposes each of them in unit of a bit and outputs them with decreasing priority. Parameters or bits of low priority are omitted by the bit stealer, depending on the transmission bit rate. One the receiver side the parameters are extracted and the speech signals are decoded on the basis of arrangement type identification codes transmitted together with the coded data.