Chihiro Takeshima

Pitch shifts in scale alternated wavelet sequences and the prediction by auditory image model and spectral temporal receptive field

SAWSs are acoustic stimuli in which an impulse response of vocal tract and its scaled version are alternately placed in the time domain at a constant periodic rate. When the scale factor is close to unity (1.0), the perceived pitch corresponded to the original periodicity. As the difference in the scaling became large, the pitch tended to be matched to what corresponds to be lower than the original by an octave. One of the characteristics of this pitch shift was that the pitch chroma did not change. This sort of pitch continuum could not be realized by changing the fundamental frequency of harmonic complex tones, but could be realized by attenuating the odd harmonics of them. Two auditory models were used to predict this pitch shift phenomenon, i.e., AIM by Pattersons group; STRF by Shmmas group. Both models could predict the pitch shift by an octave, but AIM predicted the pitch ambiguity better than STRF. While it is easy to find the secondary local peak of periodicity besides the primary peak, the pe...

Pitch perception for sequences of pulses alternating different resonance scales.

Vowels are produced as a sequence of vocal tract impulse responses which are periodically excited by glottal pulses. Each impulse response reflects the shape and the size of the vocal tract. The size, i.e., the resonance scale, is kept almost constant in the normal speech sounds. It has been found that we can sensitively “hear” such size variations in speech and that the size can be a cue to identify the sound source, i.e., the talker. What do we perceive if impulse responses with two different resonance scales are alternatively repeated. Do we perceive an intermediate sized talker speaking the same pitch as the original? Do we perceive two speech streams of the different sized talkers with the half pitch? By answering these questions, we could clarify the auditory mechanisms for detecting size, periodicity, and resonance. The test stimuli were synthesized whose consecutive glottal pulses were alternated between two resonance scales with several degrees. Pitch matching experiments were performed where the...

Temporal characteristics of extraction of size information in speech sounds

We can identify vowels pronounced by speakers with any size vocal tract. Together, we can discriminate the different sizes of vocal tracts. To simulate these abilities, a computational model has been proposed in which size information is extracted and separated from the shape information. It is important to investigate temporal characteristics of the size extraction process. Experiments were performed for listeners to detect the size modulation in vowel sequences. All the sequences had six segments. Each segment contained one of three Japanese vowels: ‘‘a,’’ ‘‘i,’’ and ‘‘u.’’ Size modulation was applied by dilating or compressing the frequency axis of continuous, STRAIGHT spectra. Modulation was achieved by changing the dilation/compression factor in sinusoidal functions. The original F0 pattern of the base sequence, except for warping of the time axis, was used for all stimuli. The minimum modulation depth at which listeners were able to detect the existence of modulation was measured as a function of th...

Interaction between the tonotopic and periodic information in recognition of ‘‘melodic’’ contours

It is known that the tonotopic and periodic aspects of sounds are separated in the auditory sense. To investigate the manner of representing these two aspects, a psychological experiment was designed. The stimuli are sequences of vowels, each of which changes its fundamental frequency and ‘‘size,’’ i.e., the centroid of its formants, using a STRAIGHT vocoder. Participants are asked to respond if the transient patterns of the stimuli are ‘‘clockwise’’ or ‘‘counterclockwise’’ in a two‐dimensional space, one of whose coordinates is the periodicity and the other of which is the size. Results indicate that the judgment was not so easy as in the case in which one would judge the transition patterns of points in a two‐dimensional visual space. This inference implies that the two attributes might be represented as values on two completely independent scales, but not like the values in two coordinates of a single point.