Showing papers by "Unto K. Laine published in 2013"

Time-frequency integration characteristics of hearing are optimized for perception of speech-like acoustic patterns.

Abstract: Several psychoacoustic phenomena such as loudness perception, absolute thresholds of hearing, and perceptual grouping in time are affected by temporal integration of the signal in the auditory system Similarly, the frequency resolution of the hearing system, often expressed in terms of critical bands, implies signal integration across neighboring frequencies Although progress has been made in understanding the neurophysiological mechanisms behind these processes, the underlying reasons for the observed integration characteristics have remained poorly understood The current work proposes that the temporal and spectral integration are a result of a system optimized for pattern detection from ecologically relevant acoustic inputs This argument is supported by a simulation where the average time-frequency structure of speech that is derived from a large set of speech signals shows a good match to the time-frequency characteristics of the human auditory system The results also suggest that the observed integration characteristics are learnable from acoustic inputs of the auditory environment using a Hebbian-like learning rule

Attention based temporal filtering of sensory signals for data redundancy reduction

Abstract: Since modern computational devices are required to store and process increasing amounts of data generated from various sources, efficient algorithms for identification of significant information in the data are becoming essential. Sensory recordings are one example where automatic and continuous storing and processing of large amounts of data is needed. Therefore, algorithms that can alleviate the computational load of the devices and reduce their storage requirements by removing uninformative data are important. In this work we propose a method for data reduction based on theories of human attention. The method detects temporally salient events based on the context in which they occur and retains only those sections of the input signal. The algorithm is tested as a pre-processing stage in a weakly supervised keyword learning experiment where it is shown to significantly improve the quality of the codebooks used in the pattern discovery process.

Automatic self-supervised learning of associations between speech and text

Abstract: Discovery of statistically significant patterns from data and learning of associative links between qualitatively different data streams is becoming increasingly important in dealing with the so-called Big Data problem of the modern society. In this work, a methodological framework for automatic discovery of statistical associations between a high bit-rate and noisy sensory signal (speech) and temporally discrete categorical data with different temporal granularity (text) is presented. The proposed approach does not utilize any phonetic or linguistic knowledge in the analysis, but simply learns the meaningful units of text and speech and their mutual mappings in an unsupervised manner. The first experiments with a limited vocabulary of childdirected speech show that, after a period of learning, the method is successful in the generation of a textual representation of continuous speech.