DE69421774T2 - Method and device for speed conversion of speech - Google Patents

Description

The invention relates to a speech rate changing method, a speech rate changing device and an electronic device for modulating the rate of a voice, and more particularly, it relates to a technique effective when applied to a control technique for using such a device in conversation, etc.

As a device for assisting hearing perception in hearing-impaired people, analog hearing aids using analog circuits to process the amplitude and frequency characteristics of a voice have been conventionally used. In contrast, in recent years, intensive research and development have been made on applying digital signal processing to compensate for hearing impairment. The trend of research and development is described in detail in, for example, "Application of Digital Technique to Compensation for Hearing-Impairment", Journal of Acoustical Society of Japan, (Vol. 47, No. 10, pp. 760-765, 1991), "Speech-perception aids for hearing-impaired people: Current status and needed research", J. Acoust. Soc. Am. (90(2), Pt. 1, August 1991), etc.

In order to compensate for hearing loss, amplification of the amplitude of a speech signal and compression of the dynamic range for each frequency are generally carried out according to the hearing characteristics of a user. In a conventional analog hearing aid, such a process is carried out by an analog circuit. On the other hand, in digital hearing aids developed in recent years, this process is carried out by software such as a digital filter or etc., so that adaptation to the hearing characteristics of a user can be carried out in more detail.

In the above trend, in recent years, an attempt has been made to change only the speech speed by digital signal processing without any change in the signal period of a voice, thereby performing higher quality hearing aid measures covering the entire range of a hearing system, including a reduction in the speech processing speed. Such a speech speed changing technique is described in detail in, for example, "Development of Portable DSP System for performing Speech Processing for the Aged", Technical Research Report of Institute of Electronics, Information and Communication Engineers of Japan (Vol. 92, No. 207 SP92-54), "High-Quality Real-Time Speech Speed Conversion System", ditto (SP92-55), etc.

In the above-mentioned conventional techniques, a voice transmitted via television/radio or the like or a voice recorded on a tape recorder or the like was used as the voice to be subjected to voice speed change. That is, the object of voice speed change was only a sound signal that was unilaterally input to a listener.

However, taking into account the fact that conventional hearing aids can be used without distinguishing the type of input voice, it is preferable that the speech speed changing device can also use voices other than the above-mentioned voices as the input voice. In particular, when a speaker's voice can be heard slowly in a conversation, the device can be used not only in the case of hearing aids for the elderly or hearing-impaired, but also in the case of hearing aids in conversation in a foreign language that is unfamiliar to people without hearing impairments, etc.

Speech speed changing methods and devices with the features in the preamble of claim 1 or claim 2 are known from DE-A-42 27 826.

SUMMARY OF THE INVENTION

It is an object of the invention to provide a technique that can expand the application range of speech rate change .

This object is achieved by the invention characterized in claims 1 and 2, wherein when a time delay occurs due to speed change or a repeat operation, catching up of speech is carried out while reproducing stored speech. This allows a reduction in the operation time and improves the handling characteristics of a speech speed changing device.

Preferred embodiments of the invention are set out in the dependent claims.

According to claim 3, data is stored frame by frame, so that the write/read efficiency can be improved.

According to claim 4, a decision regarding processes for expanding and reducing a waveform, a process for eliminating a silent part, etc. in the speech speed changing process is carried out based on a comparison between the power in a frame and a threshold value, and the threshold value is depending on the volume of the input speech. Accordingly, the speech speed change process can be carried out according to the environmental conditions during use.

According to claim 5, in the speech speed changing device, there are provided a speech speed selection switch for selecting the speech speed and a means for changing the speech speed to the speech speed selected by the speech speed selection switch. Accordingly, the speed of the speech to be heard can be selected by the listener according to his will.

According to claim 6, in the speech speed changing device, means (AV controller) for controlling an audio/video device is provided. Accordingly, a series of operations in which a signal for interrupting the reproduction operation of an external device is output to temporarily interrupt the input of speech to the speech speed changing device when the memory capacity is insufficient and in which the output of the pause signal is stopped to restart the input of speech from the external device when there is some free space in the memory are repeated regardless of the expansion/reduction rate in the speech speed changing. As a result, speech speed changing can be continued for a long time.

According to claims 7 to 9, in the speech speed changing device, a repeat switch and a means for repeating reproduced speech in a period in which the repeat switch is turned on are provided. Accordingly, speech speed change can be carried out on the repeated speech.

According to claims 10 and 11, a catching-up device is provided so that expanding the application range of the speech speed changing device, shortening the operation time, improving the handling properties, etc. can be achieved.

According to claims 12 and 13, at least one of a speech speed change switch, a speech speed selection switch, a repeat switch and a reset switch is provided in a peripheral portion on a side surface of the speech speed change device to enable easy handling. Accordingly, expansion of the application range of the speech speed change device, shortening of the operation time, improvement of handling properties, etc. can be achieved.

According to claims 14 to 19, the speech speed changing means is provided as software executed by a digital signal processor having an input terminal for receiving an interrupt request signal from the outside, so that control of the speech speed changing process or switching of the speech speed changing rate based on the speech speed changing switch is supplied to the digital signal processor via the input terminal for the interrupt request signal.

According to claim 20, the microphone does not pick up a clicking sound of each switch, so that a loud noise when handling a switch can be prevented.

According to claim 21, the switches have respective surfaces that are provided with different grips. designed to be recognized without sight so that handling properties can be improved.

According to claim 22, a device for preventing the rustling of accompaniment in contact with the microphone is provided, so that the penetration of noise can be reduced.

According to claim 23, a display device is provided at a predetermined position of the speech speed changing device so that the amount of time delay from real time can be visually displayed. Accordingly, shortening of the operation time, improvement of handling properties, etc. can be achieved.

According to claim 24, a ring buffer is used as storage means, and there is provided means for handling a time delay by a counter indicating the time delay at the ring buffer. Accordingly, the repeat process, the catch-up process, etc. can be easily carried out.

According to claim 25, in addition to the pass-through mode, a standby mode is available so that a reduction in the electrical energy consumed can be achieved.

According to claim 26, there is provided an electric supply switch which is operated in three stages, namely an ON stage, an OFF stage and an ON-OFF intermediate stage, so as to provide an analog pass-through mode. Accordingly, a reduction in electric power can be achieved.

According to claim 27, the speech speed changing device is arranged between the handset of a telephone and the Telephone casing. Accordingly, the language to be subjected to the speech rate change can be selected by the listener without any interference with the listener's own speech.

Furthermore, the voice can be heard in the telephone at a slow speech rate without any change in the characteristics of the speaker's voice.

According to claim 28, the voice rate changing device is provided in a telephone exchange system. Accordingly, the voice to be subjected to voice rate change can be selected by the listener without any disturbance to the listener's own speech.

Still further advantages of the invention will become apparent to those skilled in the art upon reading and understanding the following detailed description of the preferred and alternative embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 is a block diagram schematically showing the structure of internal circuits according to the invention;

Fig. 2 is a graph for explaining a speech rate changing process as carried out within a DSP according to the invention;

Fig. 3 is a graph for explaining the concept of a threshold process according to the invention;

Fig. 4 is a view showing the use form of a speech speed changing device according to the present invention;

Fig. 5 is a flow chart showing the control procedure of the speech speed changing device according to the present invention;

Fig. 6 is a front view as seen from the front side of the speech speed changing device according to the present invention;

Fig. 7 is a rear view as seen from the back of the speech speed changing device according to the present invention;

Fig. 8 is a plan view as seen from the top of the speech speed changing device according to the present invention;

Fig. 9 is a left view as seen from the left side of the speech speed changing device according to the present invention;

Fig. 10 is a right view as seen from the right side of the speech speed changing device according to the invention;

Fig. 11 is a block diagram showing the functional structure of the speech rate changing device according to the present invention;

Figs. 12A and 12B are typical diagrams for explaining a compression process in a speech compression section according to the present invention;

Fig. 13 is a flowchart showing the procedure of a main process according to the invention;

Fig. 14 is a flow chart continued from the flow chart of Fig. 13;

Fig. 15 is a state transition view showing a typical transition between respective modes according to the invention;

Fig. 16 is a flow chart showing the procedure of a read pointer reset routine according to the invention;

Fig. 17 is a flowchart showing the procedure of a waveform expansion/reduction process for one frame according to the invention;

Fig. 18 is a flow chart continued from the flow chart of Fig. 17;

Fig. 19 is a flowchart showing the procedure of a parameter setting process according to the invention;

Fig. 20 is a diagram for explaining the data compression process according to the invention;

Fig. 21 is a diagram for explaining the data compression process according to the invention;

Fig. 22 is a diagram for explaining the data compression process according to the invention;

Fig. 23 is a flow chart showing the procedure for the overall operation of the speech speed changing apparatus provided with a continuously operating speech speed changing device according to the invention is;

Fig. 24 is a flow chart continuing the flow chart of Fig. 23;

Fig. 25 is a flow chart showing the procedure for the overall operation of the speech rate changing apparatus provided with a continuous speech rate changing means according to the present invention;

Fig. 26 is a flow chart continuing the flow chart of Fig. 25;

Fig. 27 is a typical view for explaining an acceleration switch used in the continuous speech speed changing device according to the present invention;

Fig. 28 is a block diagram showing the functional structure of the speech rate changing device provided with an AV controller according to the present invention;

Fig. 29 is a diagram for explaining the operation of the AV control device according to the invention;

Fig. 30 is a flowchart showing the procedure of a main process in the speech speed changing apparatus provided with an AV controller according to the invention;

Fig. 31 is a flow chart continuing from the flow chart of Fig. 30;

Fig. 32A to 32C are diagrams for explaining the arrangement operation of the microphone in the speech speed changing device according to the invention;

Fig. 33 is a view showing the structure of a modified example according to the invention;

Fig. 34 is a diagram for explaining a time delay display means in the speech speed changing device according to the invention;

Fig. 35 is a diagram for explaining an electric supply device in the speech speed changing device according to the present invention;

Fig. 36 is a diagram for explaining an embodiment in which the speech speed changing device according to the present invention is applied to a telephone; and

Fig. 37 is a diagram for explaining an embodiment in which the speech speed changing device according to the present invention is applied to a building-related broadcast system.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

Embodiments of the invention are described in detail below with reference to the drawings.

In all drawings for explaining the embodiments, parts with the same function are identified by the same reference numeral in order to avoid a repeated description.

Fig. 1 is a block diagram schematically showing the structure of internal circuits according to the invention. The reference numeral 1 denotes a DSP (digital signal processor); 11 software for executing a speech rate changing process; 12 a serial port; 13 a connector for an external interrupt flag; 14 a flag register; 2 a memory (output buffer); 3 a selector switch; 4 a PTL (push-to-listen) switch; 5 an A/D converter; 6 a D/A converter; 7 a low-pass filter; 8 a low-pass filter; 9 an analog amplifier; 10 an analog amplifier; 321 a microphone; and 325 a two-ear headphone (earphone).

In a speech speed changing device according to this embodiment, as shown in Fig. 1, speech is input to the microphone 321 and output as a speech signal (electrical signal). This speech signal is input to the A/D converter 5 via the amplifier 10 and the low-pass filter 7, in which the speech signal is converted from an analog value to a digital value at preset time intervals.

The voice signal converted into a digital value in the above-described manner is input to the DSP 1. Then, the voice speed changing process for the voice signal is carried out by the software 11 in the DSP 1. The PTL switch 4 is connected to the external interrupt flag terminal 13 provided in the DSP 1 so that the state of the PTL switch 4 is shown as the numerical value of the flag register 14 provided inside the PSP 1 to correspond to this terminal 13. In the software 11 of the DSP 1, a judgment is made according to the numerical value of the flag register 14 as to whether or not the voice speed changing process is to be carried out.

The digital voice data subjected to the voice speed change process is input into the output buffer 2. The D/A converter 6 converts the data of the output buffer 2 from a digital value to an analog value at preset time intervals. The analog signal obtained by this conversion is input to the analog amplifier 9 via the low-pass filter 8 and output from the two-ear earphone 325 as speech in accordance with the listener's preferred amplitude for the speech signal.

In this embodiment, two types of switches are provided for the PTL switch 4. One is a switch in which the power conduction is performed as long as the key is operated. The other is a switch in which the power conduction state is maintained even though the key holding state is released. The former is used in the case of conversation, whereas the latter is used in the case of continuous speech rate change for one-way speech such as speech from the radio, which is the conventional use, and the like. Furthermore, in this embodiment, both the selection switch 3 and the PTL switch 4 are connected to the external interrupt flag terminal 13 provided in the DSP 1. The numerical value in the flag register 14 is changed by the switching of the selection switch 3, so that the software 11 changes the expansion rate in the speech rate change process according to this numerical value.

Fig. 2 is a view for explaining the speech speed changing process carried out in the DSP 1 of this embodiment. The speech speed changing process in this embodiment is a method of detecting the signal period (fundamental period) of a speech signal and expanding the length of a waveform with the detected signal period as a unit, whereby a speech data set is on the order of several tens of milliseconds (- (hereinafter referred to as frame) is made into a unit for one process. Accordingly, input buffers of at least two frames in length are provided inside the DSP 1 so that while the data from the A/D converter is input into one buffer, the data stored in the other buffer is processed (pipeline process). After processing, the data is stored in the output buffer 2 having a sufficiently large storage capacity. The procedure of processing the data in each frame is as follows.

First of all, (a) a signal period extraction process (- not shown) is applied to the head part of the frame, thereby extracting the signal period of that part.

(b) Then, the data thus acquired are transferred to the output buffer 2 at one signal period.

(c) Then the data of the length of two signal periods are multiplied by a window function that changes from 0 to 1 and a window function that changes from 1 to 0.

However, the data-in positions from which the multiplication with the window functions is started are shifted by one signal period. Then, the data as the respective results of the multiplication with the window functions are added to each other to generate a reproduced wave pattern with the time length of two signal periods and to arrange the patterns behind the data of one signal period that was previously transmitted.

(d) Then, again in the state where a position two signal periods away from the data-on position previously subjected to the signal period extraction process subjected to the test is present on the head, so that the signal period detection is carried out at that position. Since the signal period of speech generally varies, a signal period different from the previously detected signal period is obtained in the second detection.

(e) Data having a length of n signal periods is transferred to the output buffer with the signal period length obtained by this final signal period detection as a unit.

The above procedure from (a) to (e) is repeated over the entire frame.

Since the signal period length depends on the input speech, the number of repetitions in one frame is not constant. Furthermore, different expansion rates are realized by changing the value of n in the above-mentioned step (e). For example, data of three signal periods in the input buffers in the state n = 1 are generated into data of four signal periods, so the expansion rate is 4/3 = 1.33 times. Similarly, the expansion rate in the state n = 0 becomes 1.50 times, and in the state n = 2 it becomes 1.25 times.

Furthermore, in this embodiment, the above-mentioned speech speed changing process shown in Fig. 2 is not always applied to all frames, but the above-mentioned process of Fig. 2 is applied only when the calculated average power of each frame exceeds a preset threshold Th. Data in a frame with a power not exceeding the threshold Th is therefore sent to the output buffer in its original state. Fig. 3 shows a concept for this threshold process.

In Fig. 3, the part where the power of each frame exceeds the threshold Th is represented as the expansion period. Since the front and back parts of the speech signal are not processed but are output in their original state in this thresholding process, there is an advantage that the speech characteristic existing in the front and back parts of the speech, such as the consonant characteristic, is not destroyed.

Furthermore, in this embodiment, a second threshold value To is provided in the threshold process for the average power of each frame, as shown in Fig. 3. Accordingly, when a frame having a power below the second threshold value To continues for a time not less than one second, the frame having the power below the threshold value To is continuously processed for a time not less than one second so as not to be output. Accordingly, a reduction in the amount of data stored in the output buffer is achieved.

In Fig. 3, this non-output part is shown as the omission period. From the output buffer 2, data is output one by one to the D/A converter 6 at regular time intervals in parallel with the one-shot writing of the data processed in terms of the speech speed change process. Addresses in the output buffer 2 are entered in the form of a ring so that the last address is continued with the first address.

Accordingly, in this ring-shaped address space, an operation is carried out such that an address pointer Po pointing to a data value to be supplied to the D/A converter is placed behind an address pointer Pi pointing to the destination of writing a data value subjected to the speech speed change process. In this embodiment, Pi overtakes the pointer Po sooner or later because the speed of Pi is higher than that of Po. At this time, information stored in the output buffer 2 is not output but rewritten.

Accordingly, the time from the start of the speech speed changing process to this state becomes the time period of the input speech that can be tackled by the speech speed changing process of this embodiment. The reduction of the data amount based on the above-mentioned threshold value To has the effect of making the time period that can be tackled long.

Furthermore, the signal processing method in the speech speed change process explained above with reference to Figs. 2 and 3 was reported in "Technical Research of Institute of Electronics, Information and Communication Engineers of Japan, SP92-150 (1993-03) or "Discussion of Speech Speed Conversion Method using Portable DSP System", Proceedings of Acoustical Society of Japan (March 1993), 1-7-6.

Fig. 4 is a view showing the use form of the speech speed changing device according to this embodiment. Although Fig. 4 shows the case that the PTL switch 4 is provided on the top of the device, it is a matter of course that its arrangement position may be replaced by another position. On the other hand, the selection switch 3 for changing the expansion rate in the speech speed change is provided on one side of the PTL switch 4. Since both the selection switch 3 and the PTL switch 4 are provided so that the State of the selector switch 3 is observable by the software in the DSP 1 through the connection to the DSP 1 for the external interrupt flag, the value of n in the above-mentioned speech rate changing process changes according to the state of the selector switch 3 when the PTL switch 4 is operated. The expansion rate can be changed for each language by alternately operating the PTL switch 4 and the selector switch 3.

Fig. 5 shows the above control procedure as a flow chart. In the speech rate changing process, for example, a frame with a time period on the order of several tens of milliseconds is used as a processing unit. The A/D conversion and the D/A conversion are processes that are executed at regular intervals with a time period smaller than this, for example, at regular intervals with a time period on the order of several tens of microseconds. As shown in Fig. 5, the A/D conversion, the D/A conversion and their related process are realized as an interrupt process. While the speech rate changing process and a process for waiting for an interrupt are executed, the interrupt process is executed in accordance with an interrupt signal from the serial port to which the A/D converter and the D/A converter are connected.

As is obvious from the above description, the speech speed changing device according to this embodiment can be used not only for speech supplied unilaterally to a listener such as radio broadcast speech but also in the situation of conversation so that the listener can select the speech subjected to the speech speed change without any disturbance of his own speech. Furthermore, the speech speed changing device of this embodiment can be used for to compensate for a speech hearing impairment such as is observed in the elderly or the like. Furthermore, the device can of course be used even in a situation where a listener without hearing difficulties hears a foreign language with which he is unfamiliar.

Referring to Figs. 6 to 10, the external structure of the speech speed changing device according to the present invention is given below. Fig. 6 is a front view; Fig. 7 is a rear view; Fig. 8 is a top plan view; Fig. 9 is a left view from the left, and Fig. 10 is a right view from the right.

In Figs. 6 to 10, reference numeral 101 denotes the casing of the speech speed changing device; 102 a rear cover; 103 a finger stop hole; 104 a slow switch (slow key); 105 a repeat switch (repeat key); 106 a reset switch (reset key); 321 a microphone; 108 a volume control; 109 an electric power switch; 110 an earphone terminal; 111 an external input terminal; 112 an AV control terminal; and 113 a speech speed changeover switch (speech speed setting switch).

As shown in Figs. 6 to 10, in the speech speed changing device of this embodiment, the slow switch 104, the repeat switch 105 and the reset switch 106 are provided at positions where the body 101 of the speech speed changing device is easily operated with one hand, for example, in the front upper part, and the speech speed changeover switch 113 is visible in the right view.

The button of the above-mentioned slow switch 104 is designed to be larger than the other buttons because the frequency of its operation is higher. Furthermore, since continuously operating the slow button is tiring, this button is designed to be locked. For example, the following types are used: (1) a sliding lock type in which the button locks when it is depressed and moved in the transverse direction; (2) a double-click type in which the button locks when it is double-clicked; (3) a type in which the locking state of the button is released when the reset button is operated, etc.

The above-mentioned speech speed changeover switch (speech speed setting switch) 113 is arranged close to a region that allows operation thereof with the same finger, so that this switch and the slow switch 104 can be operated alternately.

In addition to the position in the above embodiment, a ring switch, a slide switch, etc. may be provided to simplify operation.

The above-mentioned volume control 104 is also arranged in an area that allows operation with the same finger, so that adjustment is easy so that listening at an appropriate volume is always possible.

Furthermore, for the above-mentioned switches with high frequency of use such as the slow switch 104, the repeat switch 105, the reset switch 106, the speech speed changeover switch 113, etc., a switch with a soft touch characteristic is preferably used so that the microphone 321 does not pick up a click sound of the switch. For example, a switch is used using an electrically conductive rubber or the like.

Furthermore, the external appearances of the above respective switches are preferably formed with surface states that impart different gripping feelings in order to identify the switches by their type without looking at them.

When the above-mentioned finger stop hole 103 is opened, the structure is such that some switches such as the voice speed selection switch at a repeat timing, etc., are recognizable.

The internal circuit structure of the speech speed changing device in this embodiment is designed to be identical to the above-mentioned circuit structure shown in Fig. 1.

As the PTL switch 4 in the previous embodiment, the slow switch 104, the repeat switch 105, the reset switch 106, etc. as described above are used. Furthermore, as the select switch 3 in the previous embodiment, the voice speed changeover switch (voice speed setting switch) 113 is used. Furthermore, the voice speed changeover switch (voice speed setting switch) 113 is connected to the external interrupt flag terminal 13 provided in the DSP 1. The numerical value of the flag register 14 changes by switching the voice speed changeover switch 113, so that the software changes the expansion rate in the voice speed changing process according to this numerical value.

Fig. 11 is a block diagram showing the functional structure of the speech speed changing device in this embodiment, wherein reference numeral 21 denotes speech input devices; 22 input buffer; 23 a central processing unit (CPU); 24 a ring buffer memory (corresponding to memory 2 in Fig. 1); 25 a function selection device; 26 output buffer and 27 voice output devices.

When the components of this embodiment are made corresponding to those of the previous embodiment, the voice input devices 21 are constituted by the microphone 321, the analog amplifier 10, the low-pass filter 7 and the A/D converter 5 in Fig. 1.

The above-mentioned input buffers 22 serve to maintain a voice signal converted into a digital signal by the above-mentioned voice input devices 21, and have a size sufficient to maintain data of the length of one frame, which is the unit for subsequent signal processing. These input buffers 22 may be realized by allocating a part of addresses of the ring buffer memory 24 (which corresponds to the memory 2 in Fig. 1).

The above-mentioned central processing unit (CPU) 23, which corresponds to the part of the software executed in the DSP 1 shown in Fig. 1, has an encoder 23A, a silent part eliminating process 23B, a decoder 23C, a waveform manipulation process (speech speed changing process) 23D, and a controller 23E.

The above-mentioned function selection device 25, which corresponds to the one formed by the switches 3 and 4 and the terminal 13 for the external interrupt flag as shown in Fig. 1, consists of the slow switch 104, the repeat switch 105, the reset switch 106, the speech speed switch 113 etc., as described above.

The above-mentioned output buffers 26 serving to hold resultant data processed by the above-mentioned waveform manipulation process 23D are in practice provided in two, and each of them has a size sufficient to store data of the length of one frame expanded by the waveform manipulation. In the previous embodiment, two input buffers are provided so that a pipeline process is realized by alternately using them, whereas in this embodiment, a pipeline process is realized by alternately using two output buffers in the same manner as in the previous embodiment.

That is, while the waveform manipulation process is being carried out for one frame so that the result of the process is maintained in one output buffer, the result of the waveform manipulation process obtained in the previous cycle is output from the other output buffer via the voice output devices 27. These output buffers 26 can be realized by allocating a part of addresses of the ring buffer memory 24 (which corresponds to the memory 2 in Fig. 1).

The input of data into the input buffers 22 and the output of data from the output buffers 26 are carried out at intervals of the sampling rate of the A/D converter 5 and the D/A converter 6 in the same manner as in the previous embodiment. The process carried out by the DSP 1 therefore consists of a waveform manipulation process for each frame and an interrupt process carried out at sampling intervals. That is, the interrupt process is carried out several times while the waveform manipulation is carried out. run manipulation process is applied to data of one frame length, so that the two processes appear to be running simultaneously.

As the above-mentioned ring buffer memory 24, a memory of a well-known type in which write/read operations are carried out for each frame is used. Details of this are described below.

(Write process)

As shown in Fig. 11, voice data inputted through the voice input devices 21 are maintained in the input buffers 22. The input buffers 22 have sufficient capacity to maintain a number of data corresponding to one frame, so that a code length of 16 bits is allocated per one data, and the input buffers 22 are realized by allocating a part of addresses for the memory 2 shown in Fig. 1.

The controller 23E shown in Fig. 11 monitors the state of these input buffers 22, and it transmits voice data of one frame length to the encoder 23A whenever the input buffers 22 are filled with data of one frame length.

In the encoder 23A, the input speech data of one frame length is subjected to an information compression process so that the data which is the result of the compression process is maintained in the ring buffer memory 24. For this compression process, several methods are considered. An example is a differential data storage method as shown in Figs. 12A and 12B. Figs. 12A and 12B are typical diagrams for explaining the compression process in the encoder. 23A of this embodiment.

In this compression process, the "difference from the previous data" is continuously calculated from the initial data of each frame. In Fig. 12A, this difference data is represented as Δ1, Δ2, .... The output data of the compression process is data obtained by arranging the above-mentioned difference data Δ1, Δ2, ... with a code length of 8 bits per data after dividing the initial data of the frame into upper 8 bits and lower 8 bits. One data of the input data has a digital code length of 16 bits. However, in the case of an input signal such as a voice signal which changes sufficiently slowly compared with the sampling interval, the difference from the previous sample is not so large that this difference cannot be adequately expressed with a code length of 8 bits, which is half, as shown in Fig. 12B. The capacity of the data after the compression process is therefore approximately half the capacity of the data before the compression process, but nothing is missing in the content as long as the difference in the middle of the process does not become too large to be expressed with a code length of 8 bits.

When stored in the ring buffer memory 24, the data compressed to half the capacity for each frame are arranged in the ring buffer memory 24 in such a way that their chronological sequence is maintained.

In addition, a frame header is added to the beginning of the compressed data of each frame to indicate a break point between frames. In the compression process section, the calculation of the sum of the absolute values of all the data in the frame and also the above-mentioned compression process in Fig. 12 are carried out, and at the same time the process of recording the corresponding result as a performance value for this frame in the above-mentioned frame header is performed.

The determination of a frame to be subjected to a waveform expansion/reduction process is carried out based on a comparison between the power of the frame and the threshold value Th. Further, the process of removing the silent part is carried out based on the comparison between the power of the frame and the threshold value To.

It is preferable that these thresholds are not used as fixed values, but that they are changed depending on the volume of the input speech. For example, between the case of use in a quiet room and the case of use in a situation with large background noise, the speech rate change obviously cannot be carried out well unless these thresholds are well set.

In a specific implementation method, the maximum and minimum values of the frame power in the past period of several seconds are stored so that the above-mentioned threshold values are determined based on these values. For example, if these threshold values are to be changed at intervals of five seconds when the condition exists that the time period of one frame is 50 milliseconds (ms), the process of changing the threshold value Th can be carried out once every 100 frames are processed.

As described above, the power of each frame is always calculated for all input signals, regardless of whether information compression is performed by the encoder in Fig. 11 is performed for each frame so that the associated information is recorded in the frame header and maintained in the ring buffer 24.

In this frame power calculation, the maximum frame power Pmax and the minimum frame power Pmin are compared with each other so that they are updated as needed. If the maximum frame power Pmax and the minimum frame power Pmin are provided to be reset at intervals of five seconds (100 frames), the maximum frame power and the minimum frame power in the past five-second period can always be maintained.

For example, when calculating the thresholds, Th and To are set to 10% and 5% of the difference between the maximum frame power Pmax and the minimum frame power Pmin, respectively. This is given by the following expressions (1) and (2):

Th = Pmax - Pmin 0.10 + Pmin (1)

To = Pmax - Pmin 0.05 + Pmin (2)

Although the method for maintaining the output data in the ring buffer memory 24 of this embodiment has been described above, details for removing a silent part will be described below.

As explained in the previous embodiment with reference to Fig. 3, the silent part removing function is for removing a silent part (duration during which the power is lower than the speech part/silent part threshold value To) that lasts for a time not less than one second.

The silent part eliminating process is executed by the silent part eliminating process 23B shown in Fig. 11. This silent part eliminating process is a process independent of a later-described process executed for each frame (hereinafter referred to as a main process), so that the process is executed after the main process for one frame is completed. In Fig. 14, the process is executed between the judgment "Time delay = 0?" (S143) and the judgment "Voltage ON?" (S144) (which is not shown).

In the silent part eliminating process 23B, data accumulated in the input buffers 22 are added up at intervals of a predetermined unit (e.g., 1/4 frame) to calculate the power, so that the silent part eliminating process is started when the power "crosses the speech part/silent part threshold value upwards." This is because the time of termination of the silent part is the time of power change from a small value to a large value, and, at any time other than that time, no judgment can be made as to whether or not the silent part continuing up to that point is longer than one second.

When the silent part elimination process is started, the first step is to retrieve the frame header in the ring buffer 24 backwards. Compressed data in the ring buffer 24 is compressed for each frame and the power of the frame is recorded in the frame header as described above. When a frame with a power below To continues for a time not less than one second, silent part elimination is activated and the entry pointer of the ring buffer 24 is reset to the time at which the silent part portion lasted for one second. The input of the next compressed data value is recorded so that overwriting occurs from the deferred point in time. Accordingly, a silent portion which has lasted for a time not less than one second immediately before the current point in time is always eliminated.

(Reading process)

The main process in the device of this embodiment described later is carried out for each frame. Therefore, the waveform manipulation process 23D shown in Fig. 11 maintains the currently processed frame data so that the reading from the ring buffer memory 24 is carried out collectively for each frame. That is, since the addressing of the ring buffer memory 24 can be easily carried out by a process of simply increasing the address bit by bit when data is collectively taken out, this case shows better efficiency than a case of taking out data one by one.

Since the data stored in the ring buffer memory 24 is compressed data as described above, it is necessary that this compression be decoded to the original data before the waveform manipulation process. For this purpose, the decoder 23C shown in Fig. 11 is provided. First, the leading two 8-bit data values are arranged as upper/lower of the 16 bits for the compressed data of one frame as input information to generate a leading data value. Then, the third data value of the compressed data is added to the leading data value to restore the second data value. Then, the next data value of the compressed data is added to the second data value to restore the third data value as Then, the process of adding the compressed data to the already restored data is repeated continuously to restore all the data for the frame.

The basic operation of the speech speed changing device in this embodiment will be briefly described below.

As shown in Fig. 11, the speech converted into a digital signal by the speech input devices 21 is first input to the input buffers 22. The speech signal read from the input buffers 22 is input to the encoder 23A included in the CPU 23 of the DSP 1 (Fig. 1), subjected to the data compression process, and stored in the ring buffer memory 24. The above-mentioned speech signal is also supplied to the silent part removing process 23B so that this silent part removing process is applied to the data stored in the ring buffer memory 24 as needed.

The data of the voice signal stored in the ring buffer memory 24 is supplied to the decoder 23C frame by frame so that the compressed voice data is decoded by the decoder 23C and input to the waveform manipulation process (voice speed change process) 23D. In the waveform manipulation process (voice speed change process) 23D, a voice speed change or the like is carried out based on the condition set by the function selector 25. The digital voice data subjected to the voice speed change process or the like is maintained in the output buffers 26. The data of the output buffers 26 are read out so that the voice subjected to the voice speed change process or the like is output from the speech output devices 27.

That is, the data of the output buffers 26 are read out so that the data is converted from a digital value to an analog value at intervals of a set time by the D/A converter 6 as shown in Fig. 1. The analog signal obtained by this conversion is input to the analog amplifier 9 via the low-pass filter 8 and output from the two-ear earphone 325 as speech at the listener's preferred amplitude for the speech signal.

Referring to Figs. 11, 13 and 14, the process executed for each frame in this embodiment (hereinafter referred to as the main process) will be described below.

Figs. 13 and 14 are flowcharts showing the procedure of the main process in this embodiment.

As shown in Fig. 13, in the main process according to this embodiment, a "run-up" step is carried out when the power is turned on (S131). That is, immediately after the power of the electric power supply is turned on, the data stored in the output buffers 26 are indeterminate. Therefore, immediately after the power of the electric power supply is turned on, data unrelated to speech may be output. If the data from the speech output devices 27 are output in cadence, they may form noises of a very high level. To prevent this, in this embodiment, the data values in the output buffers are adjusted by executing the run-up step so that the output signal of the speech output devices gradually increases for a predetermined time after the power of the electric power supply is turned on, regardless of the data in the output buffers 26. More specifically, whenever a data value is transferred from the output buffers to the D/A converter, this data value is multiplied by a coefficient, so that this function is realized by changing the value of the coefficient with the passage of time. This operation is carried out by the controller 23E shown in Fig. 11.

After that, the process for the "pass-through mode" is started. In the process of the pass-through mode, a "read pointer matching" step is first executed (S132). This read pointer matching process is a process in which, when data is input from the voice input devices 21, the same data is input to the output buffers 26 immediately after it is input to the input buffers 22. This process is realized by causing the value of the input pointer pointing to an input address of the memory to be matched with the value of the output pointer pointing to an output data address in the memory immediately after the data is input to the input buffers 22. In Fig. 11, this process is executed by the controller 23E.

After the read pointer agreement, in the pass mode, the actuation states (ON states) of the slow switch 104 and the repeat switch 105 are checked (S133 and S134). If both switches are in the non-actuation state (-OFF states), this situation returns the routine to the previous read pointer agreement step (S132) so that the pass mode is continued. Accordingly, in the interrupt process that occurs while the pass mode is continued, input data is always output intact so that the same speech as the input speech is output from the speech output devices 27.

In Fig. 11, the above-mentioned respective switches such as the slow switch 104, the repeat switch 105 and the reset switch 106 are included in the function selector 25, and their states are checked by the controller 23E.

When the repeat switch 105 is operated (turned ON) in the above-mentioned pass mode, a separately provided repeat flag (not shown) is set from 0 to 1, and a "read pointer reset" routine is executed (S135). Fig. 16 shows a flowchart of the internal procedure of this read pointer reset routine. The explanation of Fig. 16 will be given later.

When the slow switch 104 is operated in the above-mentioned pass mode, the location in the routine goes to the routine for "setting a parameter in the expansion process" (S136) as shown in Fig. 13. Fig. 17 shows a flowchart for the internal procedure of this routine. An explanation of Fig. 17 will be given later.

After the parameter setting in the expansion process is carried out, the waveform expansion/reduction process for one frame is carried out (S137). Figs. 18 and 19 show flowcharts for the internal procedure of this waveform expansion process for one frame. The explanation of Figs. 18 and 19 will be given later.

After the above process for one frame is completed, the location in the routine proceeds to a step of checking the states of the respective switches as to whether a respective switch is actuated or not. Thereafter, since the process for one frame is within half the period of one frame, the process is completed in the order of several tens of milliseconds (ms). On the other hand, switching elements such that the operation states are maintained for a time not shorter than the operation period, no matter how short it is, when the respective switches (keys) are operated by a user are used in this device. Accordingly, the location in the routine can be adjusted to a desired operation with such a time delay that the user does not get a feeling of slow response, as long as the operation states of the switches are checked every time the process is carried out for one frame.

First, it is checked whether the reset switch 106 is pressed or not (S138). If the reset switch 106 is pressed (if yes in S138), the current mode at that time is forcibly changed to the pass-through mode.

When the reset switch 106 is not depressed (in the case of NO in S138), it is checked whether the slow switch 104 is depressed or not (S139) as shown in Fig. 14. When the slow switch 104 is depressed (in the case of ON: in the case of YES in S139), the location in the routine returns to the routine in which a parameter in the expansion process is set so that the waveform expansion process is continuously applied to the next frame. When the slow switch 104 is continuously depressed, the location of the routine continuously loops in this loop. Furthermore, even when the slow switch 104 is continuously depressed in repeat playback and in catch-up playback, the location in the routine continuously loops in this loop.

If the slow switch 104 is open (in the case of OFF: in the case of No in S139), the place in the routine proceeds to the next judgment on the repeatedly operated state (S140). The case according to which the depression of the repeat switch 105 is recognized at this time is either that in which "the repeat switch is operated at the time of repeat reproduction" or that in which "the repeat switch is operated at catch-up reproduction". In either case, the place in the routine branches to the read pointer reset routine so that the repeat reproduction is started from the silent part near a position retarded by about five seconds from the current position of the output pointer of the ring buffer memory 24.

When the slow switch 104 is open and the repeat switch 105 is not operated, the place in the routine proceeds to the following repeat end judgment (S141). The repeat operation is continued until the output pointer returns to the output pointer position at which the pass mode was switched to the repeat operation by operating the repeat switch 105. That is, when this judgment shows that the repeat mode is currently being used and the position of the output pointer has not yet returned to the output pointer position at which the repeat was started, a processing loop is formed such that the place in the routine returns to the above-mentioned waveform expansion/reduction process for one frame. The following process is a process for catch-up reproduction.

After repeat playback has ended, or after slow playback has ended, the location in the routine moves on to catch-up playback. Catch-up playback be means an operation in which a time delay from the real time caused by the repetitive or slow playback is made up by fast playback, which is realized by repeating the waveform reduction process for one frame. In the process in this section, a parameter setting for the waveform reduction process for the catch-up playback (S142) was made.

The amount of delay from real time increases when the repeat button is pressed or when the waveform expansion process is executed. On the other hand, it decreases when the waveform reduction process is executed.

However, the process for increasing/decreasing this delay amount is not shown in Figs. 18 and 19, which show the procedure (flowchart) for the waveform expansion/reduction process for one frame, which will be described later.

A judgment is made as to whether a value of the time delay from the real time is present or absent (S143). If it is still the case that a value of the time delay is present, a processing loop is formed so that the catch-up reproduction is continued. That is, by this judgment, an operation is realized in which the catch-up reproduction is continued until the value of the time delay becomes zero.

On the other hand, in the main process described above, the time delay from real time caused by voice rate change or repeat operation is managed using a counter as a "delay measure".

Although the time delay from real time can also be managed as the difference between the position in the ring buffer 24 where the currently sampled data is input and the position in the ring buffer 24 where the position of output data is input, i.e., the difference between addresses pointed to by two pointers, the invention uses the management method using the delay amount counter as described above. This is because it may be possible that the amount of delay cannot be correctly reflected by the address difference between the input and output pointers in the ring buffer 24.

For example, if it is assumed that the memory address space allocated to the ring buffer 24 runs from address 0 to address 1000, the ring buffer 24 is implemented by manipulating the memory address space in such a way that the program "jumps to address 0 after address 1000." Therefore, if the input and output pointers are across this dividing point between addresses, the amount of data in between cannot be easily represented by simply taking the difference between address values. To find the amount of data between these pointers by address calculation, an address value calculation is required that uses a complicated classification that takes into account the histories of the two pointers up to their current positions.

However, in the speech speed changing device according to the present invention, whenever reading/writing of data is carried out with respect to the ring buffer 24, the value of the delay amount counter changes to thereby manage the time delay amount to prevent an increase in the To prevent excessive processing due to complicated address calculation.

The above main process is in the form of an infinite loop in which the above process is repeated until the electric supply switch is turned off (S144).

When the electric power switch is turned off, the process is not stopped suddenly, but continues for a predetermined time before it is terminated (muting) (P145). During this time, a process is carried out such that the volume of the output voice is gradually reduced.

More specifically, in an interrupt process, in the same manner as the start-up process which is the first step, whenever a data value is transferred from the output buffers 26 to the D/A converter 6 as shown in Fig. 1, this data value is multiplied by a coefficient so that this function is realized by changing the value of this coefficient with the passage of time. This process is carried out by the controller 23E shown in Fig. 11.

Fig. 15 is a state transition view typically showing the transition between respective modes in this embodiment described above. From Fig. 15, the manner of mode switching based on a switching operation is clearly seen. Furthermore, the standby mode in Fig. 15 will be described in detail later.

The details of processing operations in the respective routines of this embodiment described above will be described below.

Fig. 16 is a flowchart showing the procedure of the read pointer reset routine.

The read pointer reset routine in this embodiment is a special method for changing the value of the output pointer pointing to the position of the data to be read out from the ring buffer 24, which method is necessary for realizing a repeat function.

As shown in Fig. 16, first, the position of the output pointer at the current time is set to Pout (S161). Then, the delay amount from the real time at the current time is set to D (S162). A judgment is made as to whether the delay amount at the current time is already so large that it exceeds the size of the ring buffer memory 24 if it were to increase by another five seconds (B5) (S163). If, as a result of the judgment, a decision is made that the delay amount exceeds the size of the ring buffer memory 24 (if Yes in S163), this routine is terminated without any change in Pout and d (S69 and S170).

If the delay amount can be increased by five seconds (B5) (case No in S163), the pointer is set back by five seconds (-B5) and the delay amount is increased by five seconds (+B5) (S164).

Then a process of searching backwards for a silent part is started so that the start of the repetition is a pause in speech. First, a data value is searched backwards with respect to the position to which Pout in the ring buffer 24, to thereby calculate the power for one frame (S165).

At this time, when the output pointer is moved back by one frame (F) (-F), the delay amount is further increased by one frame (+F). At this time, a judgment is made as to whether or not the total delay amount exceeds the size of the ring buffer memory when the delay amount is further increased by one frame (S166), and if as a result of the judgment, a decision is made that the total delay amount exceeds the size of the ring buffer memory (case Yes in S166), this search for a silent part is terminated and Pout and D are set to the output pointer value and the delay amount at this time, respectively (S169 and S170), whereupon this routine is terminated.

If the total delay amount does not exceed the size of the ring buffer memory (case No in S166) even though the output pointer has been retracted by one frame, Pout is retracted by the length of one frame and the delay amount D is increased by one frame (S67), whereupon the power W calculated for one frame and the speech part/silent part threshold value are compared with each other (S168). If the power W for a frame is less than this threshold value, a decision is made that there is a speech pause near this frame (case No in S168), and Pout and D are set to the output pointer value and the delay amount, respectively, at this time (S169 and S170), whereupon this routine is terminated. If the power W in a frame is greater than this threshold value (case Yes in S168), the pointer is further retracted by one frame and the search for a silent part is continued to detect a silent part in the same manner as described above, but the search is continued until the delay amount exceeds the size of the ring buffer memory. Thus, the output pointer reset process is completed at the time the repeat switch 105 is operated.

Figs. 17 and 18 are flowcharts illustrating the procedure of the waveform expansion/reduction process for one frame in this embodiment.

In the waveform expansion/reduction process for a frame according to this embodiment, first, as shown in Figs. 17 and 18, the power of the current data of a frame is calculated (S171). Then, this power value P is compared with the threshold value Th (S172). A frame with a higher power than the threshold value Th is subjected to the following process. Data in a frame with a lower power than the threshold value Th may be output intact to be transferred to the ring buffer 24 (S173), or it may be subjected to a consonant emphasis process and then transferred to the output buffers 26. Whether or not to perform consonant emphasis is determined by the state of the mode switch, which is one of the hidden switches.

As a specific method for realizing the consonant emphasis process, for example, consider a method in which a frame with a lower power than the threshold Th immediately before a frame with a higher power than the threshold Th is regarded as a consonant and the data values in that frame are emphasized.

In the case of a frame with a performance above the threshold Th in the above performance evaluation (case Yes in S172), the number of data in a frame is first men is stored in a variable Z indicating the amount of data not yet processed (S174), and then the signal period extraction process is carried out from the frame beginning (S175). As the signal period extraction process, various methods are considered. For example, the signal period length at the frame beginning is extracted based on a well-known algorithm using autocorrelation.

Then, the amount of data corresponding to twice the signal period length thus extracted is compared with the amount of data not yet processed (S176), and then, if the amount Z of data not yet processed is smaller than the amount of data which is twice the signal period extracted, this process is terminated.

When the amount of data not yet processed Z is equal to or larger than the amount of data twice the extracted signal period (case of Yes in S176), a pre-transfer process is carried out (S178). The pre-transfer process means a process in which a part of the input data is transferred intact to the output buffer 26 before a reproduction wave pattern insertion process described later is carried out. The pre-transfer process corresponds to part (b) in Fig. 2. The number of data to be transferred by the pre-transfer process is set with the signal period as a unit, but the number thereof varies depending on the waveform expansion/reduction rate. The number Npf is set by a parameter setting routine (S177) described later with reference to Fig. 19. After the pre-transmission process has been executed (S178), the amount Z of unprocessed data is reduced by the number of transmitted data (S179).

Then, the position of application of a Δ window function for generating a playback wave pattern is determined (S180) depending on another parameter Ptri set in the parameter setting routine shown in Fig. 19. The difference between expansion and reduction is only in the position in the current wave at which the window function is applied when a playback wave pattern is generated using the Δ window function.

That is, in the case of waveform expansion as shown in Fig. 2, the Δ window function is applied so that a waveform of two signal periods in length is generated from the waveform of one signal period in length (S181). In contrast, in the case of waveform reduction as shown in Figs. 20 to 22, the Δ window function is applied so that a waveform of two signal periods in length is generated from a waveform of three or four signal periods in length. The amount of delay from the real time changes by the insertion of the reproduction wave pattern (which is not shown, however).

After the playback wave pattern insertion process, the amount Z of data not yet processed is reduced by the number of data thus processed (S182).

Then, the signal period extraction process is carried out again (S183). This is a process adapted to the fact that the signal period of human speech always varies, and the deviation between the current signal period length and the signal period length for processing is reduced when the signal period is extracted again, thus preventing an increase in the distortion of the waveform after the expansion/reduction.

Then, as shown in Fig. 18, a data amount twice as large as the newly extracted signal period is compared with the number of data not yet processed (S184). If there is no longer a data amount of the length of two signal periods (case No in S184), this process is immediately terminated.

When an amount of data exceeding the length of two signal periods remains (if Yes in S184), a post-transmission process is carried out. The post-transmission process means a process similar to the pre-transmission process, and it corresponds to the section (e) in Fig. 2 in the previous embodiment. The number of data to be transmitted by the post-transmission process is set with the signal period as the unit, but the number thereof varies depending on the waveform expansion/reduction rate. The number Npf is set by the parameter setting routine (S185) described later with reference to Fig. 19. After the pre-transmission process (S186), the amount Z of data not yet processed is reduced by the number of transmitted data (S187).

The above procedure is repeated continuously until it is stopped based on the comparison between the amount of data of the length of two signal periods and the amount of data not yet processed, this comparison being carried out twice in the middle of this procedure.

Fig. 19 is a flowchart showing the procedure of the parameter setting routine for parameter setting for the expansion process in this embodiment.

In practice, the parameter setting routine shown in Fig. 19 is used twice in the main process shown in Figs. 13 and 14. Once it is used immediately before the above waveform expansion/reduction routine for one frame, and the other time it is used in a "parameter setting process for the reduction process" after the detection of the repeat end.

Here, the waveform reduction process is a process for realizing the "catching-up process (fast-listening process)" which is connected after the slow listening or the repetition. When the generation of a reproduction wave pattern is carried out using the Δ window function as carried out in the waveform expansion process, while the position subjected to the window function is shifted in the direction backward to the case of expansion, a waveform reduction is obtained.

In Fig. 19, a first distinction is made between expansion and reduction (S191). This distinguishes one of the two cases mentioned above from the other.

In the case of the parameter setting for the expansion process, after this discrimination, the position of the speech rate selection switch is checked (S192), the expansion rate e is set according to the position of the switch (S193), the positions of the parameters Npf and Npr used in the waveform expansion process are set according to the expansion rate e, and the parameter Ptri indicating the position of the start of the weighted summation with respect to the Δ window as performed in the waveform expansion process is set, whereupon this routine is terminated.

On the other hand, in the case of parameter setting for the reduction process, the right flow in Fig. 19 is carried out. First, the position of the catch-up mode switch (the one of the hidden switches) is checked (S196), and it is checked whether the catch-up mode is set to "skip", "fast listening" or "simple" (Mcat) (S197 and S198).

When the catch-up mode (Mcat) is set to "jump", it is not actually used to "catch up", but actually jumps at the exact moment when the hold on the slow switch (slow key) is released (S199). More specifically, a branching process for forcibly returning to the pass-through mode is performed in this section.

When the catch-up mode switch (Mcat) is set to "single", the reduction rate s is set to the single (S200), and the location in the routine proceeds to step 202.

When the catch-up mode switch (Mcat) is not set to "simple", the reduction rate s is set by the middle procedure in Fig. 19 in the case of the catch-up mode (S201), the values of the parameters Npf and Npr used in the waveform reduction process are set in accordance with the reduction rate s (S202), and further, the parameter Ptri indicating the position from which the weighted summation with respect to the Δ window performed in the waveform reduction process starts is set (S203), after which this routine is terminated.

Figs. 23 and 24 are flow charts showing the procedure of the overall operation of a speech rate changing device according to the present invention, which is provided with a continuous speech rate changing device.

As shown in Figs. 23 and 24 in this embodiment, continuous speech rate change in the speech rate changing device provided with continuous speech rate change means is essentially an operation in which the operation of the slow switch (slow key) 104 continues so that slow playback continues. However, the time delay quickly accumulates when waveform expansion continues at a constant waveform expansion rate so that the amount of delay from real time eventually exceeds the capacity of the ring buffer 24, making it impossible to continue slow listening any further.

Therefore, the continuous speech rate change facility is provided to mix a waveform expansion period and an inverse waveform reduction period during slow playback so that the delay from real time does not increase rapidly.

Although several methods are considered as means for switching to the continuous speech rate change mode, it is quite easy to see that a clear distinction must be made between the case where the operation of the slow switch is simply continued for a fairly long time and the case where entry into the continuous speech rate change mode is intended. Accordingly, this switching is carried out, for example, by using switching parts by which the slow switch is locked when it is double-clicked (operated twice within a short time interval) or when it is moved sideways while being depressed.

The respective steps in the flowcharts shown in Figs. 23 and 24 of this embodiment are quite the same as those in the procedure of the main process described above with reference to Figs. 13 and 14.

In the continuous speech rate change measure according to this embodiment, in a step S231 in Figs. 23 and 24, it is checked whether a continuous speech rate change process is intended or not (S231). If the continuous speech rate change process is intended (if Yes in S231), the waveform expansion/reduction process is carried out for one frame (S232). Then, a judgement is made as to whether the reset switch 106 is operated (turned on) or not (S233). If the reset switch 106 is not operated (turned off), a count-up of one frame is carried out (S234), and a judgement is made as to whether an expansion period is provided or not (S235). If an expansion period is provided (if Yes in S235), the place in the routine goes back to the step S232. If no expansion period is intended (case No in S235), a parameter for the reduction process is set (S236). Then, it is checked whether the delay amount is zero or not (S237). If the delay amount is zero (case Yes in S237), the place in the routine returns to step S232. If the delay amount is not zero (case No in S237), a parameter for the expansion process is set (S238) and the frame counter is reset (S239), whereupon the place in the routine returns to step 232 so that the continuous speech speed change process is repeated. If no continuous speech speed change process is intended in the above-mentioned step 231 (case No in 231), the mode switched to the above-mentioned main process routine (pass-through mode).

That is, the continuous speech speed change measure in this embodiment is a method in which slow playback and catch-up playback are alternately repeated at intervals of a preset time. According to this method, catching up to real time is always enabled at intervals of a predetermined time. Management of switching between waveform expansion and waveform reduction is carried out based on the count value for the number of frames. For example, when the expansion process is completed for a number of frames corresponding to about five seconds, the reduction process is then repeatedly carried out, and when the delay amount reaches zero, the frame count value is reset to zero and the expansion process is repeated again.

Further, exit from the continuous speech rate change mode is achieved by depressing the reset switch 106 to reset the mode to the continuous mode.

Next, reference is made to Figs. 25 and 26, which are flow charts showing the procedure of the overall operation of a speech rate changing device provided with a different continuous speech rate changing mechanism from the embodiment shown in Figs. 23 and 24.

The continuous speech rate change in the speech rate changing device provided with a continuous speech rate change measure according to this embodiment is a pre Process in which waveform expansion is applied for a high power frame and waveform reduction is applied for a low power frame.

In the continuous speech rate change measure according to this embodiment, it is checked in a step S251 in Figs. 25 and 26 whether the continuous speech rate change process is intended or not. If the continuous speech rate change process is intended (case of YES in S251), a judgment is made as to whether the reset switch 106 is operated (turned on) or not (S252). If the reset switch 106 is not operated (turned off), the power for one frame is calculated (S253). Then, it is checked whether the calculated power for one frame is higher than the threshold value Th or not (S254). If the power calculated for one frame is lower than the threshold value Th (case of NO in S251), a parameter for the reduction process is set (S256), and the place in the routine returns to the step S257. When the power calculated for one frame is higher than the threshold Th (case of Yes in S254), a parameter for the expansion process is set (S255), and the waveform expansion/reduction process for one frame is carried out (S257), whereupon the place in the routine returns to step S252 so that the continuous speech rate change process is repeated. When the continuous speech rate change process is not intended in the above-mentioned step S251 (case of No in S251), the mode goes to the above-mentioned main process routine (pass-through mode). That is, when the continuous speech rate change mode is entered, the power for one frame is calculated so that for each frame, based on the comparison between the power of one frame and the threshold Th, either expansion or reduction is applied. Exiting the continuous speech rate change mode is achieved by operating the reset switch 106.

In this embodiment, the speech is made slow or fast depending on its performance.

In normal conversation, there is generally a tendency that an important part to be communicated to a listener is spoken louder, but a part that is not so important is spoken softer. Accordingly, the speech rate control in this embodiment is characterized in that output speech closer to natural speech is achieved.

However, the probability of occurrence of a high-performance part and the probability of occurrence of a low-performance part are not always equal to each other, so that catching up to real time at intervals of a predetermined time, as in the case of the previous embodiment of Figs. 23 and 24, is not always ensured.

Further, as a method for instructing the user to enter the continuous voice speed change mode, a method in which the slow switch (slow key) 104 is operated and then slid sideways to be locked thereby, a method in which the slow switch (slow key) 104 is double-clicked (operated twice consecutively with a short time interval), etc. are considered. When these methods are used, the respective intentions of "executing slow playback" and "continuing" the operation by operating the slow switch (slow key) 104 can be expressed by varying the manner of operating the same key, so that an operating system can be provided which is more intuitive and easier to understand than if a separate key for continuous speech rate change is provided.

The previous embodiments have been described on the assumption that the waveform "expansion rate" in the case of "slow" reproduction based on the waveform expansion process is determined by the setting of the "speech speed setting switch" provided on the device, and that a "default value" determined in advance (in the program) is used as the waveform "reduction rate" in the case of "fast" reproduction based on the waveform reduction process.

However, the function of "allowing free coming and going on the speech time axis" provided by this device can be used more intuitively by the user if an "acceleration type switch" shown in Fig. 27 is provided so that the waveform expansion/reduction rate is changed by this switch.

When the acceleration type switch is set to the center, the pass-through mode is executed in the above embodiments. When the slide switch is pulled forward, the waveform expansion process is executed so that "slow playback" is executed with a delay from real time. Then, when the slide switch is pushed backward, the waveform reduction process is used conversely so that fast playback is executed (until the delay from real time reaches zero).

During this time, the controller changes the waveform expansion/reduction rate according to the distance from the center of the slide switch. However, as is clear from the explanation of the above embodiment in Figs. 20 to 22, the expansion/reduction rate can only be set to various values that are reproducible in integer rates. In practice, therefore, the expansion/reduction rate may preferably be set so that a plurality of value levels can be selected depending on the distance from the center of the slide switch.

Furthermore, if a lever is provided such that a force for returning the lever to the center automatically acts when the user releases the finger holding force for this acceleration type switch, it is easy for the user to hold the slide switch at an intermediate position other than the center, so that an easy-to-handle operating method can be realized. Here, the generation of the force for returning this lever to the center can be achieved by two springs provided inside a switching member to thereby form a mechanical action such as an action for pulling the lever with simultaneous force from opposite sides, etc.

Next, reference is made to Fig. 28, which is a block diagram showing the functional structure of a speech rate changing device provided with an AV controller. Fig. 29 is a view for explaining the operation of the AV controller in the embodiment of Fig. 28. Reference is made to Figs. 30 and 31, which are flowcharts showing the operation in the main process in the speech rate changing device provided with an AV controller according to this embodiment.

As shown in Fig. 28, the speech speed changing device provided with an AV controller of this embodiment is provided with a functional structure in which an AV controller 28 connected to the controller 23E is added to the functional structure of the speech speed changing device of the above-mentioned embodiment shown in Fig. 11.

The above-mentioned controller 23E judges whether or not a condition for outputting an AV control signal is satisfied, and operates the AV controller 28 to start/stop the output of the AV control signal.

As shown in Fig. 29, the AV control means is a software in which the AV control signal is output when the amount of delay from the real time caused by slow or repeated playback exceeds a predetermined value (30 seconds in Fig. 29), and in which the output of the same signal is stopped when the amount of delay reaches zero by catch-up playback.

The AV control signal is taken out of this device and used to temporarily stop the reproducing operation of a recording/reproducing device such as a tape recorder, a video tape recorder or the like. By this measure, it is possible to continue slow listening of input speech which continues for a long time exceeding the capacity of the ring buffer 24 in this device.

In Fig. 30 and 31, the dashed line The line-enclosed portion is a step showing the operation of the AV control device and is added to the flowcharts in Figs. 12 and 13. In this step, a judgment is made as to whether or not the condition for outputting the AV control signal is satisfied (S301). The judgment as to the output of the AV control signal is made by judging whether or not the amount of delay from the real time in a loop in which the waveform expansion/reduction process is repeated for one frame for slow or repeated playback is more than 30 seconds (S301), and the output of the AV control signal is started when the amount of delay from the real time is more than 30 seconds (S302).

On the other hand, the process for stopping the AV control signal is executed immediately after the judgment "delay amount = 0?", which is a judgment for exiting the loop of the catch-up reproduction process (S303).

Next, reference is made to Figs. 32A, 32B and 32C, which are views for explaining the arrangement of a microphone in a speech rate changing device according to the present invention. Reference numeral 101 denotes the casing of the speech rate changing device; 321 a microphone; 322 an extendable and collapsible support for holding the microphone 321; 323 a flexible support for holding the microphone 321; and 324 an electric cable for electrically connecting the microphone 321 to the casing 101 of the speech rate changing device via a wire.

Fig. 33 is a view showing a modified example of this embodiment of Fig. 32, wherein the reference numeral 101 denotes the housing of the speech speed changing device; 104 a slow switch; 105 a repeat switch; 106 a reset switch; 321 a microphone; 324 an electric cable for electrically connecting the microphone 321 to the housing 101 of the speech speed changing device; 325 an earphone and 300 a connector element.

In the arrangement of the microphone in the speech rate changing device of this embodiment, as shown in Fig. 32A, the microphone 321 is supported by the extendable and collapsible support 322. Since the above-mentioned support of the microphone 321 makes the microphone 321 far away from the body 101 of the speech rate changing device, a rustling noise caused by clothing, such as that generated when the device body is put into a breast pocket in use, can be prevented.

Alternatively, as shown in Fig. 23B, the microphone 321 is supported by a flexible support 323. Since the microphone 321 is supported in this manner, it is separated from the voice speed changing device casing 101 and can be bent in a desired direction. Accordingly, a rustling noise caused by clothing can be prevented when the device casing is put into a breast pocket in use.

Further, as shown in Fig. 32C, the microphone 321 and the housing 101 of the voice speed changing device are electrically connected to each other via a wire (or wirelessly). The S/N ratio can be improved because the microphone 321 and the housing 101 of the voice speed changing device are electrically connected to each other via the wire (or wirelessly) as described above. which are connected so that the microphone 321 can be arranged in the vicinity of the listener independently of the housing 101 of the speech speed changing device.

Furthermore, as shown in Fig. 33, the body 101 of the speech speed changing device and the microphone 321 are electrically connected to each other via an electric cable through the earphone 325 and the connector 300. Further, operation switches such as the slow switch 104, the repeat switch 105, the reset switch 106, etc. are provided on the above-mentioned connector 300. In this way, not only can a rustling of clothes, which is generated when the device body is put into a breast pocket during use, be prevented, but also the S/N ratio and handling properties can be improved.

Next, reference is made to Fig. 34, which is a diagram for explaining a time delay display means in a speech speed changing apparatus according to another embodiment of the invention. Reference numeral 341 denotes a display part and 342 a display screen.

As shown in Fig. 34, the time delay display means in this embodiment displays how much a speaker's speech is delayed from the real speech speed in the above-mentioned slow/repeated playback. For example, if it is assumed that the human image in Fig. 34 represents a time delay of 10 seconds, the time delay from the current time is expressed in terms of the number of human images displayed. In this way, the time delay amount from the current time is visually recognizable. Accordingly, both the speaker and the Listener can easily adjust the speech rate change, so that this device is usable in such a way that it is easy to handle. The visual display of the time delay is made, for example, by providing a liquid crystal display in the center of the front of the case of the speech rate change device shown in Fig. 6 and by displaying the display screen as shown in Fig. 34 on the liquid crystal display. Further, this display part is controlled by a "liquid crystal display driver" (not shown) connected to the controller 23E in Fig. 11.

Since the time delay amount to be displayed is constantly managed by the time delay counter in the main process illustrated in Figs. 13 and 14, the numerical value of this delay amount counter can be converted at a conversion rate of 1 to 10 seconds, so that a corresponding number of human images can be displayed on the above-mentioned display. This display process is carried out by the controller 23E in Fig. 11 by means of the above-mentioned display driver, and the timing of the rewriting of the display is then sufficient as long as the rewriting is carried out every time the processing of one frame is completed. For example, this display process is carried out between steps S137 and S138 in Fig. 14.

Fig. 35 is a view for explaining an electric supply device in a speech speed changing device according to the present invention. Reference numeral 1000 denotes a device part concerning the speech speed changing device; 1, a DSP; 5, an A/D converter; 6, a D/A converter; 9, an analog amplifier; 10, an analog amplifier; 1001, an electric supply; 1002, an electric power supply line. tion and 1003 a switch.

In the speech speed changing device according to this embodiment, as shown in the state transition view of Fig. 15, a standby mode is provided besides the pass mode so that the standby mode is automatically entered when the pass mode has continued for a predetermined time. That is, when either the slow switch or the repeat switch is operated (turned on), the clock frequency is increased so that each process is executed.

Furthermore, in the pass-through mode, the DSP 1 operates with a faster clock signal, but power is wasted because no voice speed change process or the like is carried out. Therefore, in the standby mode, the operation clock for the DSP 1 is lowered so that only input/output of data is carried out, thereby reducing the consumed electric power. Furthermore, only storage in the memory is carried out. In this way, a voice storage function is realized.

Further, as shown in Fig. 35, in the analog pass-through mode, the changeover switch 103 is connected to a contact side on which the electric power supply line 1002 is cut off, and it is also connected to a contact side for directly connecting the analog amplifiers 10 and 9 so that no electric power is supplied to the DSP 1, the A/D converter 5, the D/A converter 6 and digital peripheral circuits. At this time, no storage is carried out in the memory. That is, the analog input/output systems are directly connected to easily operate as an analog amplifier. The above changeover switch is designed as a three-stage switch. namely, for an ON stage, an OFF stage and an ON-OFF intermediate stage, as shown in Fig. 35, so that the analog pass-through mode is provided.

As is apparent from the above explanation, according to this embodiment, a three-stage switch is formed having an ON stage, an OFF stage, and an ON-OFF intermediate stage so as to provide the analog pass-through mode. Accordingly, not only a reduction in electric power can be achieved, but also the range of use of the electric supply can be expanded.

With reference to Fig. 36, an embodiment in which the speech rate changing device according to the invention is applied to a telephone will now be described. The reference numeral 2000 designates the speech rate changing device according to the invention; 3000 a telephone set; 3001 a transceiver and 3002 a telephone line.

As shown in Fig. 36, the telephone in this embodiment is constituted by inserting the speech rate changing device 2000 of the present invention between the handset 3001 and the telephone set 3000. Further, in the case of a cordless handset or cordless transceiver 3001 for children, the speech rate changing device 2000 is inserted between the transceiver 3001 and the telephone set 3000 by cordless connection.

Furthermore, the speech speed changing device according to the invention can be used as a speech speed changing device in a switching system, so that it can be operated upon request by the user.

With the above structure, speech can be heard slowly over the telephone. Furthermore, since the speech is fed back to the listener's side as a pass-through speech while being heard slowly by the listener, there is no risk of difficult speaking, so that the speaker can speak normally during a telephone conversation with an elderly person or the like.

Furthermore, it is unnecessary for any A/D device to be present inside the speech speed changing device as long as the speech speed changing device is designed as a digital circuit.

Referring to Fig. 37, a description will be given of an embodiment in which the speech rate changing device according to the present invention is applied to a building-based broadcast system. Reference numeral 2000 denotes the speech rate changing device; 321 a microphone; 325 an earphone; 4003 an amplifier; and 4004 a speaker.

In the telephone of this embodiment, as shown in Fig. 37, the speech speed changing device 2000 according to the invention is inserted between the microphone 321, the earphone 325 and the amplifier 4003 for the speaker 4004.

With the above structure, the listener can hear the speech at an appropriate speech speed even if the speaker does not control the speech speed change process. For example, the listener can hear the speech at an appropriate speech speed even if the speaker is selfish and verbose at a high speech speed. ity (impulsive language).

Furthermore, the listener can hear a speaker at an appropriate speech rate even if the speaker speaks slowly.

As is apparent from the above explanation, the invention can be applied to technical fields that require speech rate variation, such as hearing aids, language learning, foreign travel, music, etc., besides telephones, telephone switching systems and building-based paging systems.

For example, in the case of products for learning languages and travelling abroad, the invention can be applied in the following cases:

(1) Recorded speech is listened to continuously and slowly.

(2) The expansion rate will be changed according to progress in the learning level.

(3) A part that is difficult to understand in normal speech; listened to repeatedly and slowly.

(4) After slow listening, speech is listened to again at its original speed.

(5) After slow repetition, the pronunciation is imitated.

(6) Imitated speech is listened to in comparison with the original language.

(7) Multiple people listen to a source simultaneously at their preferred speech speeds.

Furthermore, in combination with a digital audio device such as a tape recorder, a CD recorder, an MD recorder, etc., it is unnecessary for any A/D converter to be present in the speech speed changing device, so as long as the audio device has digital output.

Furthermore, the invention can be applied to musical purposes, as long as changes are made in the following way.

No assessment is made based on the performance of an expanded frame (as the tempo will be disrupted).

The signal period extraction range is extended compared to the case of speech.

The waveform expansion process is performed based on a signal period of fixed length. In the case of associated speech, the signal period is detected so that the detection is performed based on the detected signal period.

A foot switch is provided so that a change operation can be carried out by the foot switch. This makes it possible to control a musical instrument while playing.

Although the invention has been specifically described based on embodiments thereof, it is to be understood that it is not limited to the above-mentioned embodiments, but that many changes can be made within the scope of the claims.

In short, effects achieved by typical embodiments of the invention disclosed in this application are as follows.

(1) Since the speech speed changing device can be used not only for speech that is familiar to the listener Not only is the speech fed to the listener at a constant speed, like radio speech, but since it can also be used in conversational situations, the speech to be subjected to a speech speed change can be selected by the listener without any disruption to the listener's own speech.

Furthermore, with hearing aids, foreign language learning devices, telephones, etc., the speaker's voice can be heard at a slow speech rate without any change in the speech characteristics.

(2) Effective use of memory, a repeat function of the source language, a voice storage function, a repeat function with voice speed change, a playback function with fast listening, etc. can be provided.

(3) Since a means for changing the speech speed to a value selected by the speech speed change switch is provided, the speech speed of speech to be heard can be selected by the listener at the will of the listener.

(4) Since a means for repeating reproduced speech is provided in a period in which the repeat switch is turned on, the speech speed of repeated speech can be changed.

(5) Since a catching-up device for catching up to a position of stored information to be heard is provided in the speech speed changing device, an extension of the application range of the speech speed changing device, a shortening of the operation time, an improvement of the handling properties, etc. can be achieved.

(6) Since at least one of a speech speed change switch, a speech speed selection switch, a repeat switch and a reset switch is provided in a peripheral portion located on a side surface of the speech speed change process, and since this switch is easy to handle, expansion of the application range of the speech speed change device, shortening of the operation time, improvement of handling characteristics, etc. can be achieved.

(7) The efficiency of the speech speed change process can be improved.

(8) Since not only the setting of the waveform expansion/reduction process and the silent part elimination process within the waveform expansion/reduction process are made based on a comparison between the power of a frame and the threshold value, but also the threshold value is changed according to the volume of the input speech, the waveform expansion/reduction process can be carried out according to the environmental conditions of use.

(9) Since the microphone does not pick up the clicking sound of switches, the reproduced speech can be heard clearly.

(10) Since the switches have respective surface shapes with different grip properties to be recognized without looking, the handling properties can be improved.

(11) Since a device is provided to prevent rustling of clothing at the microphone, the ingress of noise can be reduced.

(12) Since a display device capable of visually displaying the amount of time delay from the current time is provided at a predetermined position of the speech speed changing process, a reduction in the operation time, an improvement in handling properties, etc. can be achieved.

(13) Since a ring buffer is used as a storage means, so that a means for managing the time delay by means of a counter indicating the time delay in the ring buffer is provided, complicated calculation of pointer addresses in the repeat process, the catch-up process, etc. can be easily carried out.

(14) Since a standby mode and an analogue pass-through mode are provided in addition to the pass-through mode, a reduction in the electric energy consumed can be achieved.

(15) Since the electric supply switch is in the form of three stages with an ON stage, an OFF stage and an ON-OFF intermediate stage, so as to provide an analogue pass-through mode, not only can a reduction in electric power be achieved, but also the application range of the electric supply can be expanded.

(16) Since the above-mentioned speech rate changing device is provided between the handset of a telephone and the body of the device, the language to be subjected to speech rate changing can be selected by the listener without disturbing the listener's own speech.

(17) The voice of a speaker can be heard over a telephone with a slow speech rate without any change in the speech characteristics.

(18) Since the speech rate changing facility is provided in a telephone switching system, the speech to be subjected to speech rate changing can be selected by the listener without any disturbance to the listener's own speech.

Claims (3)

1. A method for changing the speed of speech, wherein speech input is received and stored and its speed is changed without changing its signal period, and wherein a speed change of the speech input is carried out during a period determined by a listener in which a change in the speech speed is required, but not during the rest of the time, characterized in that the amount of delay from real time is adjusted in a period in which the stored speech is reproduced if the delay is caused by a change in speech speed or a repetition process. 2. Device for changing the speed of speech, with means (321) for receiving voice input; a storage device (2) for storing information reproducing the voice input; a speed change device (1) for changing the speed of the voice input; a device (325) for supplying an output of the speed change device (1) as a voice output to the ears of a listener; a speech speed change switch (113); and a device designed to output speech in the on state of the speed change switch (113) while changing the speed of the voice input and in the off state of the switch without changing the speed; characterized by a catch-up device for adjusting the degree of delay with respect to real time in a Pe interval at which the stored information is reproduced if the delay is caused by a change in speech speed or a repetition process. 3. Device according to claim 2, wherein the storage device (2) comprises a device for storing data frame by frame. 4. Apparatus according to claim 3, further comprising means for deciding in the speed change process regarding pulse diagram expansion and contraction processes based on a comparison between the energy of a frame and a threshold value provided as a variable. 5. The apparatus of claim 2, further comprising a speed selector switch (113) for selecting the speed of the speech and means for changing the speed of the speech to the speed selected by the speed selector switch. 6. Apparatus according to claim 5, further comprising means for controlling an audio/video device. 7. Apparatus according to claim 2, further comprising a repeat switch (105) and means for repeating speech reproduced in the on state of the repeat switch. 8. The apparatus of claim 7, wherein the repeating means comprises means for switching the speech back by several seconds when the repeat switch (105) is pressed once, means for generating intermittent tones several times while the speech is is switched, means for stopping the speech switching back when the speech reaches the end of a ring buffer (24), and/or means for selecting the speed during the repeat time. 9. The apparatus of claim 8, wherein the means for selecting the speed during the repeat time comprises at least two of the following modes: repeat at a default speed, slow repeat, fast repeat, gradually increasing speed repeat. 10. The apparatus of claim 2, wherein the catching up means includes means for starting a catching up operation when a slow playback mode is terminated, means for starting a repeat operation when playback is switched back to the time of starting a repeat after the repeat, means for selecting the speech speed during the catching up time, means for automatically switching the running mode to a pass-through mode for directly outputting the speech input when the catching up operation is terminated, and/or means for generating a confirmation beep when the catching up operation is terminated. 11. The device according to claim 10, wherein the means for selecting the speed during the catch-up time comprises a means that performs a jump to real time without stopping, a means that catches up with real time with fast listening, and/or a means that performs a parallel movement with a time delay. 12. Device according to claim 2, further comprising a speed selector switch, a repeat switch (105) and/or a reset switch (106), wherein the switch or switches are provided in an edge area on a side surface of the speed change device for easier handling. 13. The apparatus of claim 12, wherein the reset switch (106) includes means for stopping the repeat or catch-up operation and jumping to real time if the switch is turned on during the repeat or catch-up time and then switching the running mode to a pass-through mode. 14. The apparatus according to claim 2, wherein the speed change means is provided as software (11) executed by a digital signal processor (1) having an input terminal for receiving an interrupt request signal from the outside, so that the control of the speed change operation or the switching of the ratio of the speed change due to the speed change switch (113) is supplied to the digital signal processor (1) via the input terminal for the interrupt request signal. 15. The apparatus of claim 2 further comprising means for listening to the voice output via dual-ear headphones (325). 16. The device of claim 2 further comprising a microphone (321) for converting an acoustic signal into an electrical signal; an analog amplifier (10) for amplifying an output signal of the microphone (321); a low-pass filter (7) for removing high frequency components from the output signal of the analog amplifier (10); an analog/digital converter (5) for converting the analog output signal of the low-pass filter (7) into a digital signal; a digital signal processor (1) for carrying out the speed change process; means (113) for changing a processing parameter; a digital/analog converter (6) for converting the digital voice data into analog values; a second low-pass filter (8) for removing high frequency components from the output signal of the digital/analog converter (6); a second analog amplifier for amplifying the output signal of the second low-pass filter (8); and a headphone (325) for converting the output signal of the second analog amplifier into an acoustic signal and for supplying the acoustic signal to both ears. 17. The apparatus of claim 16, wherein the speed changing means repeatedly performs a sequence of procedures over a full frame by frame-by-frame pipelining using a plurality of input frame buffers, and wherein the sequence of procedures includes: performing a signal period extraction operation on the initial part of the frame to detect the signal period of the initial part; Transferring data of the length of a signal period thus captured to the output buffer; Multiplying data of the length of two signal periods using a window function that changes from 0 to 1 and from 1 to 0; Adding the respective data obtained by the multiplications using the window functions to generate a reproduced wave pattern with the temporal length of two signal periods; Inserting the reproduced wave pattern at the end of the preliminarily transmitted data of the length of one signal period; again carrying out a signal period detection operation at a location which is two signal periods away from the location at which the signal period extraction operation was previously carried out so that the detection of the signal period is carried out at that location, and transferring data of length n signal periods (where n is an integer) based on the signal period length achieved in the last signal period detection to the output buffers. 18. The apparatus of claim 17, wherein the speed changing operation is performed only when the average energy of data in an input frame is higher than a preset threshold, and wherein the data contained in the frame is directly transmitted to the output buffers when the average energy is below the threshold. 19. The apparatus of claim 18, wherein in the thresholding process, a second threshold is provided for the average energy of the data in the input frame when a frame having an average energy below the second threshold continues for a period of time longer than a predetermined temporal threshold, and wherein the data in a frame having an average energy below the second threshold and continuing for a period of time greater than the temporal threshold is not allowed to be transmitted to the output buffers. 20. Device according to one of claims 2 to 15, wherein the or each switch has a soft-touch characteristic, so that the microphone (321) does not pick up any clicking noise of the switch. 21. Device according to claim 20, wherein the or each switch has different surface shapes in the grip to enable identification without looking. 22. Device according to one of claims 2 to 21, further comprising a rustle suppression device which changes the distance between a microphone (321) and the device housing (101) such that the microphone (321) does not directly touch the clothing when the device housing (121) is inserted into a breast pocket during use. 23. Apparatus according to claim 2, further comprising a display device provided at a predetermined location of the speed change device for visually displaying the degree of time delay compared to real time. 24. Device according to one of claims 2 to 23, wherein a ring buffer (24) serves as the storage device (2) and the device further comprises a device for managing the time delay by means of a counter which indicates the time delay at the ring buffer (24). 25. Device according to claim 16, wherein in addition to the pass-through mode, a standby mode is provided in which the clock cycle of the processor (1) is reduced and the same process as in the pass-through mode is carried out. 26. Apparatus according to claim 2, further comprising an electrical supply switch (109) operating in three stages, namely an ON stage, an OFF stage and an ON-OFF intermediate stage, and an electrical power supply device operating in an analog pass-through mode in which analog input/output systems are short-circuited and thereby directly connected to one another to interrupt the supply of electrical power to a digital processing system between the analog input/output systems when the switch is in the intermediate position. 27. A telephone comprising between the handset and the telephone set the speech rate changing device according to claim 2. 28. A telephone line switching system comprising the speech rate changing device according to claim 2. 1) Special note on Fig. 30 and 31 The labeling in the steps not provided with a step number in Fig. 30 and 31 corresponds to the labeling in Fig. 12 and 13 respectively. In addition, the following steps are present in Fig. 30 and 31: S301 Delay ≥ 30 sec? S302 Turning on the AV control S303 Turning off AV control 2) Translation of labels with reference symbols or with step numbering 1 DSP = digital signal processor 2 storage 7 TPF 8 TPF 11 Software 14 registers 21 Voice input device 22 Input buffer 28 AV Control 101 Housing of the voice speed change device 2000 Speech speed change unit 3000 Telephone 23A Encoder 23B Processor to eliminate a silent part 23C decoder 23D Waveform Manipulation Processor 23E Control 24 ring buffers 25 Function selector 26 Output buffer 27 Speech output device S131 Boot S132 Match between read and write pointers S133 Slow switch ON? S134 Repeat switch ON? S135 Reset reading pointer S136 Set parameters in the expansion process (Slow mode) (Repeat mode) S137 Execute waveform expansion/reduction process for one frame S138 Reset switch ON? S139 Slow switch ON? S140 Repeat switch ON? S141 Repetition finished? S142 Set parameters in the reduction process (catch-up mode) S143 Time delay = 0? S144 Voltage ON? S145 Mute S161 POUT = Value of the output pointer at the current time S162 D = Time delay compared to real time at the current time S163 D + B5 > Ring buffer size? S165 W = power calculated for one frame in reverse direction S166 D + F > Ring buffer size? S168 W > Noise/silence threshold? S169 Set output pointer to POUT S170 Set time delay from real time to D S171 Calculate P = Power in current frame S172 P > threshold Th? S173 Transfer all data of the current frame to the output ring buffer S174 Storing Z = number of data in a frame S175 Extract signal period S176 Z > Data volume of two signal periods? S177 Enter the number of transmission signal periods in the parameter Npf S178 Transfer data to the ring buffer according to the signal period (pre-transmission) S179 Z = Z - (number of data values transferred) S180 Application position for triangular window function set according to parameter Ptri S181 Insert playback wave pattern S182 Z = Z - (number of processed data values) S183 Extract signal period S184 Z > Data volume of two signal periods? S185 Enter the number of transmitted signal periods in the parameter Npf S186 Data transmission according to the signal period to the ring buffer (post-transmission) S187 Z = Z - (number of data values transmitted) S191 Expansion (slow listening) or reduction (catching up)? S192 Check the position of the voice speed change switch S193 Set expansion rate e according to switch position S194 Set the number Npf of pre-transmission signal periods and and the number Npr of post-transmission signal periods according to e S195 Set triangle window function application position parameter Ptri for expansion use S196 Check the position of the catch-up mode (Mcat) switch S197 Mcat = jump? S198 Mcat = simple? S199 Set so that the pass-through mode is implemented immediately when the slow switch is switched off S200 Set reduction rate s when catching up to simple S201 Set reduction rate s to the default value S202 Set the number Npf of pre-transmission signal periods and and the number Npr of post-transmission signal periods according to s S203 Set triangular window function application position parameter Ptri for reduction use S231 Continuous speech rate change? S232 Execute waveform expansion/reduction process for one frame S233 Reset switch ON? S234 Increment number of frames S235 Expansion period? S236 Set parameters in the reduction process S237 Time delay = 0? S238 Set parameters in the expansion process S239 Reset frame counter S252 Reset switch ON? S253 Calculate the performance of a frame S254 Power > Th? S255 Set parameters in the expansion process S256 Set parameters in the reduction process S257 Execute waveform expansion/reduction process for one frame 3) Translation of labels not yet given in alphabetical order 1st pitch // 1st signal period 2nd pitch // 2nd signal period 3rd pitch // 3rd signal period 30 seconds have passed in through mode kept intact // 30 seconds have passed in through mode kept intact amplitude // Amplitude A/D input // A/D input signal AV control terminal // AV control terminal buffer // buffer catching-up is continued till real time is caught up // Repeating is continued until real time is caught up catching-up reproduction // catching-up reproduction caught up (completion of reproduction) // caught up (completion of reproduction) coincidence between read and write pointers // coincidence between read and write pointers combine weighted wave-forms // combine weighted wave-forms continuous speech speed conversion process // continuous process for changing the speed of speech conversion mode // change mode data quantity of about 5 sec. // data quantity of approximately 5 sec. data quantity of one frame // data quantity of one frame delay of 30 sec. // delay of 30 sec. duration of elimination // duration of the elimination process duration of expansion // Duration of the expansion process emphasize consonants // emphasize consonants end // end expansion // Expansion expansion waveform // expansion waveform fade in // boot up frame // frame frame mem. // frame memory input waveform // input signal waveform input waveform data // Input waveform data intensity // intensity interruption // Interrupt main program // main program mute // mute no // no off // off Oh no output waveform data // Data of the output waveform perform speech speed conversion // Perform speech speed conversion perform waveform expansion/reduction process by one frame //Perform waveform expansion/reduction process by one frame pitch detecting point // signal period detection point points // points power on // power ON? power per frame // power per frame PTL flag // PTL flag real time is caught up or reset SW is pushed down // Real time is caught up or the reset switch is pushed down reduction // reduction repeat end // repeat end repeat reproduction // repeated reproduction repeat SW is continued till the time of start of read // repeat switch is continuously activated until the time of start of reading repeat SW is pushed in catching-up // repeat switch is activated when catching-up repeat SW is pushed in repeat // Repeat switch is pushed in repeat repeat SW on? // repeat switch on? repeat terminated? // repeat terminated? reproduced wave pattern // reproduced wave pattern reproduces wave pattern inserted into 2nd pitch // reproduces wave pattern inserted into the second signal period reproduction is completed to the position where depression is made initially // reproduction is completed to the position where depression is made initially. reset // reset reset SW is pushed down // reset switch is pressed down reset SW on? // reset switch on? return // back return read pointer // reset read pointer sample value // sample value set expansion rate // Set expansion rate set interruption address // set interrupt address set parameter in expansion process // Set parameters in the expansion process set parameter in reduction process // Set parameters in reduction process slow reproduction // slow reproduction slow SW is pushed continuously // slow switch is continuously pushed slow SW is pushed in catching-up // slow switch is activated when catching up slow SW is pushed in repeat // slow switch is activated during repeat slow SW on? // slow switch ON? speech speed button off // speech speed button off speech speed button on // speech speed button on standby mode // standby mode start // Start through mode // through mode time lag = 0? // time lag = 0? time // time wait for interruption // wait for interrupt weighting function // weighting function Yes Yes