Toshiyuki Aritsuka

A portable digital speech-rate converter for hearing impairment

A real-time hand-sized portable device that slows speech speed without changing the pitch is proposed for hearing impairment. By using this device, people can listen to fast speech at a comfortable speed. A combination of solid-state memory recording and real-time digital signal processing with a single chip processor enables this unique function. A simplified pitchsynchronous, time-scale-modification algorithm is proposed to minimize the complexity of the DSP operation. Unlike the traditional algorithm, this dynamic-processing algorithm reduces distortion even when the expansion rate is only just above 1. Seven out of 10 elderly hearing-impaired listeners showed improvement in a sentence recognition test when using speech-rate conversion with the largest expansion rate, although no improvement was observed in a word recognition test. Some subjects who showed large improvement had limited auditory temporal resolution, but the correlation was not significant. The results suggest that, unlike conventional hearing aids, this device can be used to overcome the deterioration of auditory ability by improving the transfer of information from short-term (echoic) memory into a more stable memory trace in the human auditory system.

Noise reduction system using neural network

A noise reduction system used for transmission and/or recognition of speech includes a speech analyzer for analyzing a noisy speech input signal thereby converting the speech signal into feature vectors such as autocorrelation coefficients, and a neural network for receiving the feature vectors of the noisy speech signal as its input. The neural network extracts from a codebook an index of prototype vectors corresponding to a noise-free equivalent to the noisy speech input signal. Feature vectors of speech are read out from the codebook on the basis of the index delivered as an output from the neural network, thereby causing the speech input to be reproduced on the basis of the feature vectors of speech read out from the codebook.

Speech coding and decoding system with background sound reproducing function

In speech decoding, a transmission code, which includes an error correcting code added to a speech code, is received and whether or not there is a code error is detected on the basis of the error correcting code. At this time, when there is no code error or when the detected code error has been corrected, a normal speech decoding processing is executed. On the other hand, when there is a code error which is impossible to be corrected, artificially background sound corresponding to the decoded speech is generated from characteristic parameters indicating unvoiced sound in the decoded speech. The parameters are continuously extracted from the decoded speech, stored in a memory and are used to replace an erroneous portion of the speech code.

Design rule for frequency-voltage cooperative power control and its application to an MPEG-4 decoder

Frequency-voltage cooperative power control (FVC) is considered a powerful method to reduce the power consumption of a program, because it utilizes the information of software loads dynamically. The authors first show through a mathematical analysis that FVC with only two frequency-voltage sets is sufficient for current low-Vdd CPU chips. Then we show an experimental result that FVC feedback control on an MPEG-4 video decoder can reduce the power to one-fourth.

Large vocabulary speed recognition using neural-fuzzy and concept networks

An algorithm for large vocabulary speech recognition using two kinds of connectionist models is described. The first one is a phoneme recognition model which uses a method combining neural nets and fuzzy inference called neural-fuzzy. This method uses neural nets as acoustic feature detectors and fuzzy logic as a decision procedure. The other is a connected-word sequence selection method using semantic information about conceptual relationships among vocabulary words. The basic idea of this method is derived from the fact that human beings can recognize words and content precisely from the topic and/or the context even when ambiguous utterances appear in conversation. The proposed method selects only word sequences that are related to each other in meaning from the several candidates, by using excitatory and inhibitory interactions with units (words). >