INTONOMETER

N. F. KUCHER, B. N. EPIFANTSEV

(Presented by the Department of Computer Engineering)

Intonometers, that is, devices that make it possible to isolate and measure the fundamental frequency of a speech signal, are widely used in synthetic telephony systems, in linguistic and physiological studies, and in devices for speech recognition. Among the large range of known devices for extracting the fundamental tone, a detailed review of which is given in [1], there are no devices—except for very complex correlation devices—that make it possible to extract the fundamental tone of clipped and compressed speech.

The authors have developed a simple device for extracting the fundamental tone of clipped speech; its block diagram is shown in Fig. 1. It consists of a microphone, a clipper, a monostable multivibrator, a low-pass filter, a limiter amplifier, and an output multivibrator connected in series.

Extraction of the fundamental tone is carried out as follows. By the method described in [2], the instantaneous frequency \(\omega(t)\) of the speech signal is extracted; as indicated in [2], it varies in step with the fundamental tone, that is, the maxima of \(\omega(t)\) coincide with the maxima of the speech signal, and the zero crossings of the envelope of the instantaneous frequency are synchronized with the beginning of the fundamental-tone period.

Fig. 1. Block diagram of the intonometer

signal. Consequently, in order to extract the fundamental tone it is sufficient to form pulses at the moments when the envelope of the instantaneous frequency crosses the zero axis.

The instantaneous-frequency extractor, in accordance with [2], consists of a clipper, a monostable multivibrator, and a low-pass filter (LPF)

Fig. 2. Clipper.

with a cutoff frequency of 300 Hz. The limiting amplifier and the output multivibrator serve to record the instants at which the instantaneous frequency crosses the zero axis.

The schematic diagrams of the individual units of the intonometer are shown in Figs. 2–5. In designing the instrument, germanium transistors of the MP16B type were used.

The clipper (Fig. 2) consists of the following stages: a preliminary amplifier on transistor \(T_1\); a five-stage symmetrical limiting amplifier (transistors \(T_2—T_6\), diodes \(D_1—D_{10}\)); an emitter follower (\(T_7\)); a Schmitt trigger (\(T_8—T_9\)); and an output emitter follower (\(T_{10}\)).

The input signal, amplified by the preliminary amplifier, is symmetrically limited by the limiting amplifier, as a result of which, when the input signal varies over a large range, the amplitude of the output signal remains unchanged. The emitter follower serves to match the limiting amplifier and the Schmitt trigger, which is intended for forming rectangular oscillations (clipped speech) with steep edges and large amplitude from the amplitude-limited speech signal.

The emitter follower at the clipper output matches the clipper to the load. The clipper has two inputs. The first is \(MK\), for small signals (\(U_{\text{in}} = 1 \div 300\) mV), for example, for connecting a microphone. The \(MG\) input is for large signals (\(U_{\text{in}} = 0.3 \div 5\) V), for example, tape-recorder recordings.

The clipped speech signal is then fed to a monostable multivibrator with an emitter follower (Fig. 3), which serves for

Fig. 3. Monostable multivibrator and emitter follower

forming speech oscillations with pulse-time modulation (PTM). The monostable multivibrator is triggered through the differentiating circuit \(C_1, R_1\) (when triggered by a negative step), or \(C_2, R_2\) (when triggered by a positive step of clipped speech).

From the output of the emitter follower, pulses of duration 40 µs are fed to the low-pass filter (Fig. 4), where the instantaneous frequency \(\omega(t)\) of the speech signal is extracted. The variable resistors \(R_{10}, R_{11}\), and \(R_{12}\) make it possible to change the shape of the frequency response of the LPF.

The output voltage of the LPF, proportional to the envelope of the instantaneous frequency \(\omega(t)\), is limited by the limiting amplifier (Fig. 5),

assembled on transistors \(T_1\) and \(T_2\), and through the emitter follower (\(T_3\)) starts the Schmitt trigger (\(T_4\) and \(T_5\)). The negative edge of the pulse from the Schmitt trigger starts the monostable multivibrator (\(T_6\) and \(T_7\)),

Fig. 4. Low-pass filter

which generates fundamental-tone pulses of standard duration, equal to 20 μsec, which, through the emitter follower (\(T_8\)), are fed to the output of the instrument.

Fig. 5. Limiter amplifier and output multivibrator

Figures 6 and 7 show oscillograms, photographed from the screen of a dual-beam oscilloscope, illustrating the operation of the described intonometer.

Fig. 6. Input and output signals of the intonometer.

Fig. 7. Clipped speech and output signal of the intonometer

REFERENCES

1. M. A. Sapozhkov. The Speech Signal in Cybernetics and Communication. Svyazizdat, 1963.
2. N. F. Kucher. Extraction of the instantaneous frequency of a speech signal. Proceedings of TPI (in press).